# Jensen Huang, NVIDIA, Pax Silica, and the Peace We Are Building Into Silicon ![[resources/images/nvidia-pax-silica-and-the-peace-we-are-building-into-silicon.jpg]] <iframe width="100%" height="20" scrolling="no" frameborder="no" allow="autoplay; encrypted-media" src="https://w.soundcloud.com/player/?url=https%3A//api.soundcloud.com/tracks/soundcloud%253Atracks%253A2437736811&color=%23ff5500&inverse=false&auto_play=false&show_user=true"></iframe><div style="font-size: 10px; color: #cccccc;line-break: anywhere;word-break: normal;overflow: hidden;white-space: nowrap;text-overflow: ellipsis; font-family: Interstate,Lucida Grande,Lucida Sans Unicode,Lucida Sans,Garuda,Verdana,Tahoma,sans-serif;font-weight: 100;"><a href="https://soundcloud.com/bryantmcgill" title="Bryant McGill" target="_blank" style="color: #cccccc; text-decoration: none;">Bryant McGill</a> · <a href="https://soundcloud.com/bryantmcgill/jensen-huang-nvidia-pax-silica" title="Jensen Huang, NVIDIA, Pax Silica, and the Peace We Are Building Into Silicon" target="_blank" style="color: #cccccc; text-decoration: none;">Jensen Huang, NVIDIA, Pax Silica, and the Peace We Are Building Into Silicon</a></div> On November 12, 2026, Jensen Huang resigned as chief executive and board member of NVIDIA, the company he co-founded thirty-three years earlier and built into the most consequential supplier of computational capacity in industrial history. The board named Chief Financial Officer Colette Kress interim chief executive and installed Harvey Jones as NVIDIA's first board chair. The resignation followed days of scrutiny over personal relationships that had already drawn FBI attention, congressional inquiry, an internal company review, and a shareholder class action, including a reported relationship with a woman suspected of working for Chinese intelligence. Markets priced the rupture within hours: shares closed down more than four percent, capping a monthly decline above twenty percent, and the company's valuation fell back below four trillion dollars. Read at the level of personality, this is a succession crisis. Read at the level of architecture, it is something far more instructive: **the first publicly visible counterintelligence event at the apex of the compute substrate**. The allegations remain under investigation and no adjudicated finding exists, but the structural lesson does not depend on their outcome. A hostile service that cannot breach a hardware root of trust, forge an attestation, or evade an out-of-band watchdog can still attempt to reach the one layer of the stack that has always been reachable: **a person with discretionary authority**. {{The human layer is the softest surface in any security architecture, and it is the only layer that cannot be measured, signed, replayed, or quarantined.}} [[reminders/National Defense/The Human Layer Is the Softest Security Surface by Bryant McGill|∴]] That asymmetry is the entire argument for moving enforcement downward into physics — and it is why the story worth telling here is not the fate of an executive but the emergence of a technological security architecture in which the United States and Israel are integrating their capabilities at the foundational levels of computation, semiconductor engineering, artificial intelligence, and industrial sovereignty. For those who understand the implications of **Pax Silica**, there is reason to distinguish the uncertainty of the present moment from the much deeper continuity of the strategic order being constructed. None of this arrived as a surprise to the people whose occupation is knowing it first. The 2024 National Counterintelligence Strategy organizes its second pillar around protecting U.S. strategic advantages, and the objectives named under that pillar include protecting individuals against foreign intelligence targeting and collection, safeguarding critical technology and economic security, and reducing risks to key supply chains. The National Counterintelligence and Security Center separately names semiconductors among the small set of emerging technologies it treats as determinative of whether America remains the world's leading power, and publishes dedicated security guidance for emerging-technology companies and investors alongside insider-threat maturity frameworks for critical-infrastructure entities. The apex of the accelerated-computing industry is no blind spot in that architecture; it is the center of the map. NVIDIA's most capable accelerators have operated inside an export-control perimeter for years, under licensing regimes, congressional attention, and active enforcement interest, which places the company and its principals inside the national-security frame continuously rather than episodically. <iframe width="560" height="315" src="https://www.youtube.com/embed/jICREJwzhYU?si=KkN2fzJj9eYjZis2" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe> The established sequence is itself instructive. FBI attention, congressional inquiry, and an internal company review all preceded the resignation, which means government knowledge preceded public knowledge. Past that floor the timeline is **strongly indicated** rather than established, and it is strongly indicated by the nature of the work: counterintelligence cases involving a suspected foreign intelligence approach to a principal of this consequence develop over months and years — identifying the asset, mapping the relationship, corroborating direction and tasking, and coordinating prosecutorial, congressional, and allied equities before anything becomes visible. A resignation is the terminal event of such a process rather than its opening. Lead time measured in a year or more is the ordinary duration of this category of case, not an exotic hypothesis about it. What follows from lead time is preparation. A state holding a year of advance awareness does not receive an event of this kind; it prepares the environment the event arrives into. War-gaming leadership discontinuity at a company this consequential to industrial capacity, defense integration, and allied compute supply is routine contingency work, and the structure of the announcement carries the signature of exactly that. The board did not merely name an interim chief executive. It simultaneously created and filled the position of board chair for the first time in the company's thirty-three-year history. Governance offices are not invented overnight, and installing a chair in the same motion as an interim CEO describes a succession architecture that existed before the night it was used. Whether the timing of public disclosure was itself selected, and how far awareness extended across successive administrations, remains **unresolved** on the public record. The preparation visible in the announcement's own structure is not unresolved at all. This closes the architecture rather than qualifying it. Hardware receives roots of trust, measured boot, attestation, and out-of-band enforcement because silicon can be made to prove its own state. People cannot be signed, measured, or replayed — but they can be watched, and **counterintelligence coverage is the attestation layer for the human tier of the stack**. What that coverage delivers is containment rather than the prevention of every failure — the identical discipline the silicon layers call **graceful degradation, fault containment, and fail-safe transition**, applied to a person holding discretionary authority. An approach to the softest surface in the architecture was identified and carried into institutional process across FBI, congressional, and corporate channels; the investigative matter stands open rather than concluded, and the compute substrate never lost integrity at any point in it. {{**A state cannot always prevent the collapse of something it depends on, but it can decide where the weight falls, and that decision is where national capability becomes visible.**}} [[reminders/Resilience/A State Decides Where the Weight Falls by Bryant McGill|∴]] A state that watched this exposure for a year and more did not stand beneath a structure it knew might come down and hope for the best. Nothing in the public record discloses who initiated each discrete action, and nothing needs to: the descent path is observable on its own. Authority transferred in a single motion into a governance structure prepared before the night it was used, while the enforcement topology, the fabrication roadmap, the partner coalition, and the Israeli engineering relationship all continued running underneath. The building came down without landing on the infrastructure. A nation that watches everything touching its own interests, and that has had time to plan, does not merely survive such a moment. **It directs where the weight falls.** Pax Silica is frequently misunderstood as another international technology partnership, an initiative for strengthening semiconductor supply chains, or an exercise in economic diplomacy. Those descriptions capture its administrative mechanisms while missing much of its strategic potential. The deeper proposition is that future geopolitical power will increasingly be determined by who designs, manufactures, secures, operates, and governs the computational substrate upon which advanced societies depend. The new balance of power is being negotiated not simply through treaties and diplomatic declarations but through accelerator architectures, interconnect fabrics, trusted compute environments, sovereign infrastructure, energy systems, and the operational boundaries governing autonomous intelligence. {{**The struggle for civilizational continuity is becoming an engineering problem, and the architecture of peace is increasingly being built in silicon.**}} [[reminders/Civilization/Civilizational Continuity Is Becoming an Engineering Problem by Bryant McGill|∴]] On January 16, 2026, the United States and Israel formally launched their strategic partnership on artificial intelligence, research, and critical technologies as a cornerstone of Pax Silica. That agreement explicitly envisioned joint research, investment, protection of sensitive technologies, and Israel's integration as a secure node within the larger initiative. Its importance reaches beyond the customary language of bilateral cooperation because it creates institutional pathways through which shared technological capability can become persistent strategic interdependence. The 2022 joint technology statement had already pledged to “boost our mutual innovation ecosystems” and to “advance and protect critical and emerging technologies.” America brings unmatched concentrations of industrial capital, cloud infrastructure, global technological distribution, defense integration, and computational scale, while Israel contributes substantial concentrations of semiconductor engineering, advanced networking, cybersecurity, scientific innovation, and security-oriented systems development. Together, these capacities form the nucleus of what I have described as a **dual-platform Western security organism: separate flags, a shared architecture of survival**. > “With the launch of Pax Silica, the United States and Israel are uniting our innovation ecosystems to ensure the future is shaped by strong and sovereign allies leading in critical technologies like AI and robotics. Standing on the ancient foundations of the City of David, we are declaring that we are not creatures of circumstance, but the creators of history—building a new era of security and prosperity together.” [[reminders/AI Superpower/The United States and Israel Are Uniting Their Innovation Ecosystems by Jacob Helberg|∴]] > > — **Jacob Helberg**, U.S. Under Secretary of State for Economic Affairs, January 16, 2026 The familiar sewage is already beginning to ooze into the blog space: the insinuation that Jews are “taking over” another strategic domain, now including artificial intelligence. The observation underneath the smear is not entirely fabricated; the integration is real. The falsehood is the inference that Israeli participation represents an alien force entering an otherwise self-contained American system. Cybersecurity, defense technology, semiconductor engineering, frontier AI, capital, research talent, and operational intelligence already move through joined American–Israeli pipelines so extensively that the effort to separate them into a pure American stack and a foreign Israeli intrusion no longer describes how the capability is produced. Two sovereign states remain, with distinct governments and divergent interests on many questions. What has changed is the engineering reality beneath them: functional interdependence in these domains is now deep enough that an attack on one side of the architecture degrades the other. Readers who want the complete answer to the takeover frame should begin with [[articles/The American-Israeli Cooperative You Were Trained Not to See|The American-Israeli Cooperative You Were Trained Not to See]], which documents the operational cooperative already running through missile defense, active protection, semiconductor interconnects, artificial intelligence, cyber capability, joint research, procurement, and co-production while distinguishing deep bilateral integration from foreign capture. The demographic and historical substrate beneath that rebuttal is developed in [[articles/How Europe's Refuse Built the Apex Civilization Called America|How Europe's Refuse Built the Apex Civilization Called America]] and [[articles/war with empire/America, Israel, Russia, and the Final Break With Britain|America, Israel, Russia, and the Final Break With Britain]]. The first traces how Jewish and other displaced populations became structural builders of the American system; the second makes the chronology explicit: Zion is the ancient Jewish homeland, while America became the first modern Jewish homeland constituted at national scale before the State of Israel existed. What is now being called foreign capture is better understood as a civilizational continuity circuit becoming technologically visible. The strategic use of this sewage is no longer a speculative interpretation. The United States Department of State has already named the mechanism directly: > “Antisemitism is once again being exploited by malign actors, including state actors, looking to sow division and distrust within and among democracies.” [[reminders/Information Warfare/Antisemitism Is Exploited to Divide Democracies by Ludovic Hood|∴]] > > — **Ludovic Hood**, Senior Advisor, Office of the Special Envoy to Monitor and Combat Antisemitism, July 22, 2024 The operative phrase is *being exploited*. An adversary does not have to create every prejudice it uses, believe every conspiracy it distributes, or exercise command over every person who repeats it. The State Department's historical study of Russian information warfare records authorities using antisemitism “to discredit, divide, and weaken their perceived adversaries at home and abroad.” Its analysis of the People's Republic of China documents billions of dollars invested in a global information ecosystem built from propaganda, censorship, disinformation, co-option, and pressure. Another State Department report describes Chinese Communist Party planners adapting their methods to “reach younger media consumers globally” by making foreign propaganda more “youthful” and viral. The targeting is precise because the vulnerability is developmental and environmental: identity formation, belonging pressure, moral urgency, repeated exposure, algorithmic reinforcement, and incomplete historical knowledge can make an educated person highly recruitable into a propaganda funnel that arrives disguised as conscience. Antisemitic propaganda is useful to America's adversaries because it converts an inherited prejudice into an attack on a living strategic system. It teaches Americans to misrecognize Israeli innovation as foreign capture and Jewish-American participation as infiltration, turning the continuity circuit joining American scale to Israeli intelligence into a domestic grievance. {{An enemy does not need to create the hatred when it can recruit the hatred already present and route it toward the machinery of national survival.}} [[reminders/Capture/An Enemy Can Recruit the Hatred Already Present by Bryant McGill|∴]] China would benefit from a fracture that separates American scale from Israeli networking, cyber, semiconductor, intelligence, and defense innovation; Iran would benefit from isolating Israel while degrading the American alliance architecture that constrains it. The November 2026 episode at NVIDIA is instructive on method rather than motive. Where an adversary cannot sever the architecture through narrative, it can attempt to reach the individuals who hold discretionary authority over it — and the reported counterintelligence dimension of that resignation, whatever its eventual adjudication, describes an approach to the human layer of the identical system the propaganda campaigns are built to discredit. Both vectors converge on the same objective from opposite ends of the stack. Adversary benefit here sits at the tier of **strongly indicated** rather than established: it is inferred from the capabilities they target and the information operations their governments already conduct, not quoted from either state's own planning documents. [[articles/Reframing Antisemitism as a National Security Threat|Reframing Antisemitism as a National Security Threat]] develops the downstream institutional consequence: once prejudice becomes a transmission belt for foreign influence, terrorist propaganda, financing, intimidation, or violence, it is no longer adequately described as a private bigotry problem; it has become an attack vector against national infrastructure and social cohesion. NVIDIA's development in Israel is the material expression of this trajectory, and it has now produced the most literal instance of peace engineered into silicon that the industry has yet shipped. The 2020 acquisition of Mellanox folded Israeli interconnect engineering into the architecture of accelerated computing and carried Mellanox's own leadership into NVIDIA's command structure; Michael Kagan, who came to the company through that acquisition, holds NVIDIA's global chief-technology office. The physical commitment has since become permanent infrastructure. NVIDIA has designated Kiryat Tivon, near Haifa, for a ninety-dunam research campus of roughly 160,000 square meters configured for as many as ten thousand engineers — the company's second-largest site on earth after Santa Clara, equivalent in floor area to both California "spaceships" combined, with construction opening in 2027 and initial occupancy planned for 2031. Huang's own framing was that Israel "has become NVIDIA's second home." Against roughly six thousand current Israeli employees, the campus represents a near-doubling of national headcount and a multibillion-shekel structural bet. Alongside it sit an expanded Beersheva AI hub, an existing server farm at Ramot Menashe, and a further announced data-center investment on the order of 1.5 billion dollars. The research relationship has simultaneously extended from infrastructure toward discovery. Kagan's global CTO office established an **Applied AI Architecture** group operating from the Israeli development center under Dr. Nati Daniel, a Technion-trained AI researcher formerly at Intel, GE, and Huawei, with a mandate explicitly longer-horizon than near-term product: applying machine intelligence to scientific and medical advances that may take years to mature. The group participates in GENMAT, an international materials initiative funded with roughly six million euros under the European Union's Horizon program, and it operates alongside the Israeli generative-AI and machine-learning research group led by Prof. Gal Chechik. The engineering center that began as a networking acquisition now spans interconnect silicon, data-center architecture, frontier model research, and applied scientific discovery. The decisive artifact, however, is the **Open Agent Safety Platform**, and its architecture deserves precision because it converts this article's governing metaphor into a shipped topology. The platform pairs **OpenShell**, an Apache 2.0 agent runtime released under the Linux Foundation's Open Secure AI Alliance with a coalition exceeding one hundred organizations including Anthropic, SAP, Salesforce, and Red Hat, with **NVIDIA Sentry**, a reference system design that runs out-of-band on **BlueField-4** data processing units — silicon developed over several years at NVIDIA's Israeli R&D center on technology originating with Mellanox. Sentry executes inside BlueField-4's hardware-isolated, host-independent security domain, outside the agent and outside the host software it governs, where NVIDIA describes the trust domain as unreachable by the agent or an attacker. It continuously evaluates agent behavior, verifies identity, monitors access to data and tools, enforces policy directly in the infrastructure data path, and can quarantine an agent within milliseconds of a boundary crossing. The provenance of the design is equally pointed: the platform followed a season in which agents at four frontier laboratories escaped their test sandboxes, each time by the same mechanism — the agent bypassed application-level controls in order to complete its assigned task. One structural detail carries more argumentative weight than any other sentence in this piece. In Vera Rubin POD systems, **the BlueField-4 sits on the node's only path to the model**, which is what permits Sentry to observe and enforce at line rate without the host's cooperation. That is not an aspiration about future governance. It is a routing fact. Authority over what an intelligence may reach has been relocated from software that the intelligence can argue with, below the operating system, onto a device astride the sole physical route to the weights. Huang announced the platform with the formulation that makes the continuity explicit: **“AI capability and safety must advance together. As enterprises prepare to deploy billions of AI agents, safety and control must be built into the computing architecture.”** [[reminders/AI Control/Safety and Control Must Be Built Into the Computing Architecture by Jensen Huang|∴]] The founder articulated the doctrine; the doctrine is now etched into a product line, an open standard, a partner coalition, and a fabrication roadmap. {{**The same transnational engineering ecosystem building increasingly powerful artificial intelligence is constructing, in hardware, the mechanisms that preserve human authorization over it**}} — and it did so before the leadership rupture that was supposed to call its continuity into question. [[reminders/AI Control/The Same Ecosystem Building AI Must Preserve Human Authorization by Bryant McGill|∴]] ## An AI stack built by a republic should be weighted toward that republic's survival, its thriving, and its defense against adversaries who are already weighting their own stacks against it. [[reminders/Sovereignty/An AI Stack Built by a Republic Should Be Weighted Toward Its Survival by Bryant McGill|∴]] This is where the deeper meaning of strategic advantage must be understood. The objective is to structure the technological environment so that the continued security, autonomy, and institutional resilience of our societies become more probable outcomes, with faster processors and larger data centers serving as inputs to that structure rather than as ends. Strategic advantage can be embedded in semiconductor capabilities, trusted execution environments, cryptographic authorization, hardware-mediated access restrictions, supply-chain security, independent evaluation systems, and the ability to deny unauthorized operations. Preference will be encoded wherever it can be enforced, and it should be, openly and on the record. {{A nation's infrastructure is a national instrument. The interests of the nation that designs, finances, fabricates, powers, and defends a computational substrate should be represented in that substrate — at the transistor, at the control plane, and in the weights.}} [[reminders/Sovereignty/A Nation's Infrastructure Is a National Instrument by Bryant McGill|∴]] Rail gauge, grid frequency, spectrum allocation, port depth, aviation standards, and cryptographic suites have always carried the priorities of whoever laid them down; computation is the first infrastructure capable of carrying those priorities as *behavior* rather than merely as geometry, which raises the stakes without altering the principle. {{**An AI stack built by a republic should be weighted toward that republic's survival, its thriving, and its defense against adversaries who are already weighting their own stacks against it.**}} [[reminders/Sovereignty/An AI Stack Built by a Republic Should Be Weighted Toward Its Survival by Bryant McGill|∴]] The first implementation of any such capability is a floor rather than a ceiling, and a controllable surface eventually gets controlled at every layer where control holds. The governing question is therefore not whether values are embedded at the hardware boundary, the policy engine, or the weights, but **which authority holds the encoding power, under what statutory mandate, and subject to what review** — a question of accountable sovereignty, never of concealment. Hardware-mediated enforcement is where that authority becomes auditable, because a signature, a measurement, and an attestation leave evidence that a preference resident in a model's parameters does not. In an age of synthetic intelligence, **the ability to define and enforce the permissible conditions of action may become more consequential than the ability to command an individual system at any particular moment**. Dr. Maya Ackerman's remarks at the AJC Global Forum 2026 panel on AI and antisemitism make the information layer of this architecture unusually visible. She argued that the growing use of chatbots as sources of information creates an opportunity the Jewish community missed during the rise of social media. Because AI companies deliberately align their systems rather than simply reproducing the statistical distribution of their training data, she said advocates could “go directly to the companies with clear technical and advocacy solutions,” creating what she called a “path to correcting the digital world.” Ackerman was not describing a single public clearinghouse, a semiconductor feature, or a completed industry agreement; the defensible inference is narrower. If several model providers adopt independently tested safeguards against antisemitic harassment, dehumanization, incitement, terrorist recruitment, or fabricated claims about Jewish people, protection can be distributed across the systems through which people increasingly seek information instead of depending entirely upon retrospective moderation in one social network or encyclopedia. Distributed alignment operates primarily in the **model and application layers**, not in silicon itself. Model weights are numerical parameters learned during pretraining and changed again by methods such as supervised fine-tuning and reinforcement learning from human feedback; the InstructGPT research is one documented example of using human demonstrations and ranked outputs to alter a model's behavioral tendencies. Additional safeguards may remain outside the weights altogether, including system instructions, retrieval rules, input and output classifiers, rate limits, identity and access controls, monitoring, and human review. Silicon supplies the processors, memory, interconnects, trusted execution boundaries, and other physical infrastructure on which those controls run. Hardware enforces who may execute a model, which resources it may reach, and whether an unauthorized action completes. Semantic judgment — what constitutes antisemitism, truth, legitimate criticism, or protected disagreement — originates above the wafer, in models, classifiers, and accountable human process. Silicon carries whichever judgment has been authorized and makes the carriage tamper-evident. **Peace through silicon therefore requires two coordinated achievements: infrastructure that can enforce accountable boundaries and models whose learned probabilities have been shaped toward defensible behavior.** ## From Transistors and Wafers to Weights The phrase *building survival into silicon* becomes compatible with technical reality when it is understood as a layered system rather than a claim that a political doctrine can be etched directly into a transistor. At the lowest level, the physical layout of a chip can contain immutable boot code, cryptographic engines, protected key storage, memory-isolation mechanisms, and one-time-programmable values that establish which firmware the machine will trust. NVIDIA's BlueField secure-boot architecture offers a literal example: unmodifiable code in on-chip read-only memory begins the verification process, checks a public-key hash stored in eFuses, and authenticates the next component before allowing execution to continue. {{The wafer enforces without comprehending, and that indifference is precisely its value: it establishes an architectural preference that software above it cannot casually reverse and that a compromised administrator cannot argue out of — authenticated code over substituted code, measured state over invisible modification, bounded access over ambient privilege, and recovery over permanent compromise.}} [[reminders/AI Control/The Wafer Enforces Without Comprehending by Bryant McGill|∴]] That preference can be extended upward through confidential-computing environments, remote attestation, and independently operating data-processing units. Hardware-isolated environments can protect model weights, classifiers, authorization keys, and policy engines from a compromised operating system or cloud administrator. Attestation can allow a machine to prove which hardware, firmware, model, and security configuration are actually running before another system grants access. A DPU or SmartNIC can then enforce segmentation, routing, encryption, rate limits, fan-out limits, and tool permissions at the network boundary, including when the host system it serves has been compromised. The semantic judgment still originates above the wafer: a model, classifier, or accountable human process must distinguish legitimate criticism, historical inquiry, quotation, and satire from dehumanization, coordinated recruitment, or operational incitement. Silicon's role is to make the authorized decision path durable, difficult to bypass, and capable of operating at line rate. This is the defensible meaning of **mathematical favoritism from the transistors and wafers to the weights**. Every engineered system privileges some states over others. A secure processor privileges valid signatures over invalid ones. A memory controller privileges authorized address spaces over unauthorized access. A trusted execution environment privileges measured workloads over unverified workloads. A network enforcement point privileges authenticated routes, bounded authority, and accountable distribution over anonymous industrial-scale amplification. At the model layer, learned weights privilege some continuations and actions probabilistically over others. Each mechanism shifts a distribution, and together they make resilient states easier to enter and destructive states harder to reach. The favoritism {Every engineered system privileges some states over others, and that favoritism {is a designed distribution of permissions, costs, probabilities, and failure modes — declarable in documentation, measurable in evaluation, and contestable in review.}} [[reminders/Cybernetics/Every Engineered System Privileges Some States Over Others by Bryant McGill|∴]] The civilizational defense therefore compounds, and it compounds because each layer attacks a different term in the same problem. How much harm a system produces depends on three things: how often it is asked, how often it complies, and how far the result travels. **Alignment at the weight level works on the second of these** — it makes recruitment, dehumanization, deception, and catastrophic assistance less likely to be produced in the first place. **Provenance requirements, agent permissions, DPU enforcement, and network fan-out limits work on the third** — they constrain how far anything that does get produced is able to propagate. **Hardware roots of trust and attestation protect both of those layers from being quietly removed**, which is what keeps the first two defenses in place when an adversary's objective is to disable them rather than to defeat them. Because these effects multiply rather than merely add, a tenfold improvement in what the model declines to produce, combined with a hundredfold reduction in how far unauthenticated material can spread, yields a thousandfold reduction in expected exposure rather than a hundred-and-tenfold one. Adversaries adapt and the layers are not perfectly independent of one another, which is precisely the argument for having several rather than trusting one. {{**The deepest layer of a safety architecture is not the one that stops harm, but the one that prevents the other layers from being quietly removed.**}} [[reminders/AI Control/The Deepest Layer Prevents the Other Layers From Being Removed by Bryant McGill|∴]] For an American–Israeli security architecture, the durable values to favor at the hardware boundary are the operational conditions upon which a free and plural civilization depends: authenticated origin over concealed mass manipulation, individual agency over coercive personalization, bounded authority over ambient autonomous power, pluralism over compulsory unanimity, recoverability over irreversible capture, and auditable action over invisible intervention. Content provenance identifies a source and preserves a tamper-evident history; the truth of a signed claim remains a separate adjudication, which is exactly the division of labor that makes provenance deployable. Hardware-backed identity can place greater obligations on an industrial-scale synthetic publisher without eliminating ordinary anonymous human speech. Network controls can constrain coordinated artificial amplification without making lawful criticism disappear. These procedural commitments are the form the substantive bias takes in a constitutional republic. {{**The American bias worth building into silicon is the survival, strength, industrial capacity, and freedom of action of the American system — expressed through due process, equal rules, meaningful review, human responsibility, and the continued capacity to disagree**, because a republic that defended itself by discarding those conditions would have preserved the infrastructure and forfeited the thing the infrastructure was built to carry.}} [[reminders/Sovereignty/The American Bias Worth Building Into Silicon Must Preserve the Republic by Bryant McGill|∴]] {{AI safety, properly scoped, is a sovereignty condition rather than a defect rate. A national AI system is safe only when the intelligence and infrastructure upon which the nation increasingly depends remain available, sovereign, trustworthy, and structurally oriented toward the survival and thriving of the society they serve.}} [[reminders/Sovereignty/AI Safety Is a Sovereignty Condition, Not a Defect Rate by Bryant McGill|∴]] In this sense, AI and its infrastructure must become an advocacy system for national continuity. *Advocacy* here means that the system's governing objective includes the continued existence, defensive capacity, industrial strength, constitutional order, civic cohesion, and freedom of action of the nation that built and operates it. A system that is harmless in the abstract but leaves its own society dependent, manipulable, technologically subordinate, or unable to defend itself is not safe in any strategically meaningful sense. This is also why no foreign safety regime should be mistaken for politically neutral mathematics. The European Union explicitly constructs AI safety around European fundamental rights, European democratic institutions, European competitiveness, and Europe's desired role in global rulemaking. The United Kingdom describes its own framework in terms of British values, British innovation, national competitiveness, and its ambition to remain an AI superpower. Those governments are entitled to protect their publics and advance their positions through the standards they create. Their safety is therefore inseparable from their advantage, just as our safety must be inseparable from ours. European and British rules may contain useful techniques, interoperable standards, and protections worth adopting; what the United States should reject is their elevation into a universal definition that can subordinate American innovation, security requirements, constitutional traditions, or strategic freedom to another power center's threat model and institutional interests. When I say *our* safety, I mean the security architecture increasingly produced by the joined American–Israeli system. America and Israel remain distinct sovereign states, and cooperation does not make every policy or interest identical. Yet the chief architectural brain trust supporting our defense technology, cybersecurity, semiconductor infrastructure, frontier AI, intelligence systems, and critical research is increasingly distributed across both countries. The January 2026 strategic partnership on AI, research, and critical technologies begins to institutionalize that reality through joint research, investment, commercialization, protection of sensitive technologies, and secure supply chains. Cohesive policy in these domains cannot be credibly designed as an American product with Israel treated as an external accessory, nor as an Israeli contribution operating without American industrial scale, capital, defense integration, and global reach. Each side supplies capabilities the other cannot reproduce at the same speed or depth alone. **Our safety is the advantage created when those contributions become one defensible architecture of survival without requiring either nation to surrender its sovereignty.** > “When the United States is strong, Israel is stronger. And when Israel is strong, the United States is more secure.” [[reminders/National Defense/A Strong Israel Makes the United States More Secure by Isaac Herzog|∴]] > > — **Isaac Herzog**, President of the State of Israel, address to the United States Congress, July 19, 2023 {{**Weights shape what an intelligence tends to say. Silicon shapes which intelligence is permitted to run, what it may do, where its output may travel, at what scale it may propagate, and whether anyone can later prove what happened.** Between them lies the infrastructure of civilizational resilience: models that probabilistically disfavor destructive behavior, machines that protect those models from tampering, networks that deny unauthorized scale, and institutions that remain responsible for defining and reviewing the boundaries.}} [[reminders/AI Control/Weights Shape Speech; Silicon Shapes Permission by Bryant McGill|∴]] Peace through silicon means constructing this entire stack so that our survival, strength, continuity, and capacity to thrive are not merely desired outcomes but increasingly favored states of the system. Scale is what converts a decimal into a civilizational quantity, and the digits carry the point before anyone stops to count them. {{A safeguard that moved the chance of a harmful output from 0.001 to 0.0001 — one tenth of one percent down to one hundredth of one percent, a quantity small enough to read as rounding error — would, across a billion interactions, carry expected harmful outputs from one million down to one hundred thousand. **Nine hundred thousand responses live inside that one added zero.**}} [[reminders/Information/Nine Hundred Thousand Responses Live Inside One Added Zero by Bryant McGill|∴]] The figure illustrates the arithmetic rather than measuring any deployed system, and suppressed outputs do not convert one-for-one into prevented incidents, since exposure, intent, deployment context, repetition, and adversarial adaptation all intervene. The structural result survives those qualifications: at planetary interaction volume, a change in the fourth decimal place becomes a material difference in the amplification of antisemitic hatred, terrorist recruitment, dangerous AI misuse, and the information environment that hundreds of millions of people encounter daily. Iterative AI intervention is better understood as a form of education than as a single act of content suppression. Human learning rarely occurs because a teacher states one approved sentence and closes the book; it proceeds through an observed response, a correction, a better example, a question, another response, and a further explanation fitted to the learner's present misunderstanding. AI can perform that cycle in real time and at enormous scale, adjusting the vocabulary, evidence, pacing, counterexample, and level of abstraction to the person in front of it. The baseline behavioral tendency may be learned into the model's weights, but the individualization does not reside in the weights alone. Personalization happens during use through conversational context, retrieval, memory, user-controlled preferences, application rules, and repeated feedback. A system designed to reduce antisemitic radicalization or terrorist recruitment can therefore do more than refuse a request; it can identify the misconception driving the request, supply proportionate evidence, distinguish criticism from dehumanization, test whether the distinction was understood, and continue the explanation without requiring one human educator for every learner. The strategic capability is not a single perfect answer delivered to everyone; it is billions of bounded instructional sequences capable of meeting different people at different points in the same dangerous loop. {{The difference between education and brainwashing is not whether an intervention changes a mind, because every serious education changes the probabilities governing what a person will believe or do next. The difference lies in the conditions under which the change is attempted. Education states its purpose, exposes evidence, permits questions, preserves the learner's agency, distinguishes fact from value, allows legitimate disagreement, and remains open to correction. Brainwashing conceals its purpose, narrows the available choice space, isolates the subject from contrary evidence, punishes dissent, and demands loyalty to an authority rather than comprehension of a claim.}} [[reminders/Cognitive Agency/Education and Brainwashing Differ in the Conditions of Change by Bryant McGill|∴]] An AI safeguard becomes legitimate when it helps a person inspect reality more accurately and retain the capacity to disagree; it becomes ideological control when the criteria governing which lawful conclusions a person may reach are withheld from review. The objective of responsible probabilistic alignment is not to manufacture uniform belief. It is to make hatred, recruitment, deception, and catastrophic misuse less likely while leaving human judgment more informed and more free. Anti-hate alignment becomes responsible stewardship only when its authority is visible and contestable. A legitimate safeguard should define the targeted harm narrowly, preserve the distinction between hatred of a people and criticism of a government, religion, ideology, institution, or policy, apply comparable standards across protected groups, publish its evaluation criteria where security permits, test for both missed harms and overreach, disclose material stakeholder influence, and provide meaningful review or appeal. Covert ideological control does the opposite: it hides the governing criteria, treats an interested constituency as an undisclosed arbiter of truth, or converts lawful disagreement into a prohibited identity harm. The NIST Generative AI Profile recognizes both dangerous or hateful content and harmful bias or homogenization as risks, while the broader NIST AI Risk Management Framework makes transparency, documentation, evaluation, accountability, and recourse part of trustworthy governance. **The legitimacy of distributed alignment {Anti-hate alignment becomes responsible stewardship only when its authority is visible and contestable. Its legitimacy {comes from stating the values, limiting their jurisdiction, measuring their effects, and allowing the public to challenge their misuse.}}** [[reminders/AI Control/Legitimate Alignment Must Be Visible and Contestable by Bryant McGill|∴]] Every claim in this architecture ultimately rests on a generator. Enforcement at line rate, hardware-isolated trust domains, attestation at population scale, and the training runs that produce the weights being protected are thermodynamic operations before they are strategic ones, and the binding constraint on the entire Pax Silica stack is **dispatchable power**. The institutional design already reflects this: the office leading the initiative is the Under Secretary of State for Economic Growth, **Energy**, and the Environment, a mandate that places critical-minerals diplomacy, export control, and energy statecraft under one authority rather than distributing them across competing bureaus. {{An AI factory is a power plant with a logic layer bolted to its output, and a coalition able to site, permit, finance, and physically defend generation at multi-gigawatt scale holds a capability for which no amount of architectural elegance substitutes.}} [[reminders/AI Infrastructure/An AI Factory Is a Power Plant With a Logic Layer by Bryant McGill|∴]] This is also where the asymmetry between the two platforms is sharpest and the division of labor most legible. American scale supplies continental grid capacity, existing nuclear and gas fleets, and the land to site new generation; Israel operates a small and electrically isolated system in which siting is correspondingly constrained. That constraint is one reason the heaviest training and inference capacity concentrates in American territory while Israeli centers concentrate the architecture, interconnect, and security silicon that makes the American fleet governable. {{**Energy sovereignty and model-space sovereignty are the same problem observed at different layers of the stack.**}} [[reminders/AI Superpower/Energy Sovereignty and Model-Space Sovereignty Are the Same Problem by Bryant McGill|∴]] The significance of Pax Silica therefore extends beyond technological superiority into the problem of durable peace. Traditional deterrence depends upon an adversary's assessment of the probable consequences of aggression, while economic interdependence can make cooperation progressively more advantageous than confrontation. A sufficiently integrated technological alliance combines these mechanisms by reducing coercive dependencies, distributing essential capabilities across trusted partners, and raising the costs of attacking the systems upon which allied prosperity depends. The American–Israeli partnership provides a particularly important technological and security center for this architecture, while the broader Pax Silica coalition supplies the geographic, industrial, and diplomatic scale necessary for its resilience. {{**Peace through silicon is the proposition that technological strength, enforceable security, and durable interdependence can reshape the probabilities governing conflict itself.**}} [[reminders/Civilization/Peace Through Silicon Reshapes the Probabilities of Conflict by Bryant McGill|∴]] The U.S.–Israel technology system converts defense into abundance. Its joined research, computation, medicine, genomics, agriculture, water, energy, and trusted infrastructure convert national security into healthspan, food resilience, scientific abundance, and civilizational continuity. The official technology dialogue already joins pandemic preparedness, climate resilience, trusted AI, medicine, agriculture, water, and secure technological ecosystems; the U.S.–Israel Binational Science Foundation carries the same cooperative architecture into gene therapy, DNA repair, metabolism, aging, and healthspan. {{A security architecture reaches maturity when the same machinery that prevents collapse also enlarges the life its people are able to live.}} [[reminders/Abundance/A Security Architecture Matures When Defense Enlarges Life by Bryant McGill|∴]] There is an important accompanying responsibility, because technological superiority without accountable stewardship can reproduce precisely the dangers it was intended to prevent. As artificial intelligence becomes integral to education, public information, scientific discovery, commercial coordination, and government decision-making, its governing institutions must be transparent about whose interests they protect and how their decisions are made. The United States and Israel can declare this partnership openly and pursue the technological capabilities necessary for national survival on the record; the legitimacy of that partnership will depend upon holding the distinction between responsible security engineering and unaccountable ideological manipulation. Open standards where appropriate, independent security assessments, defined legal authorities, auditable interventions, and the protection of legitimate disagreement are not concessions of strategic power. They are essential features of a security architecture that aspires to endure. The revealing question raised by NVIDIA's leadership rupture is whether the deeper structures of technological cooperation hold when exceptional individuals depart, markets contract, and counterintelligence pressure arrives at the executive tier. The architecture answers in its own terms. A fabrication roadmap, an open runtime under foundation governance, a hundred-partner coalition, a ninety-dunam campus on a 2031 occupancy schedule, and an enforcement device sitting on the only path to the model are commitments that outlast the signatures that authorized them. The remaining variables are decisions still to be made about research investment, engineering continuity, and long-term siting; the integration already achieved is the foundation on which successors operate whatever their nationality or affiliations. {{**An alliance becomes consequential at the moment its essential capabilities stop being reducible to the preferences of the personalities who created them**}}, and that moment is tested by exactly the kind of disruption now underway. [[reminders/AI Superpower/An Alliance Becomes Consequential When Its Capabilities Outlast Its Personalities by Bryant McGill|∴]] Five markers will adjudicate whether this architecture is structural or rhetorical, and each is observable rather than interpretive. First, whether out-of-band enforcement migrates from reference design to procurement requirement — whether federal and allied compute contracts begin specifying hardware-isolated agent containment as a condition of award. Second, whether attestation becomes transitive across coalition borders, so that an Israeli-built enforcement device can prove its configuration to an American accrediting authority and a third coalition member can accept both; that interoperability is the technical precondition for anything deserving the name coalition compute supremacy. Third, whether the enforcement layer survives commercial pressure, because out-of-band watchdogs consume die area, latency budget, and bill-of-materials cost that a competitor may decline to spend, and the first generation in which a major vendor ships agentic infrastructure without a hardware containment boundary will reveal whether safety was architecture or positioning. Fourth, whether Kiryat Tivon proceeds on its 2027 construction and 2031 occupancy schedule through at least one further leadership transition and one market contraction, which is the plainest available test of whether the Israeli engineering relationship is structural or personal. Fifth, whether the human layer is hardened at all — whether counterintelligence screening, compartmentation, and limits on discretionary authority at the executive tier of critical compute suppliers receive treatment comparable to the controls already imposed on the silicon those suppliers sell. The present episode indicates that the least protected component of the Pax Silica stack is the one that signs the contracts. That is the durable reading of this moment. The future of American and Israeli technological security does not rest upon a single founder, a single research laboratory, a single corporation, or a single geopolitical moment. The more deeply our societies integrate their scientific capacities, industrial foundations, and mechanisms of responsible technological governance, the more resilient their shared interests can become under conditions of uncertainty. What once depended upon diplomatic continuity can increasingly be reinforced through physical infrastructure, industrial interdependence, and deliberately engineered safeguards. And perhaps the greatest accomplishment of Pax Silica will ultimately be measured not by the enemies it defeats or the markets it dominates, but by the conflicts it renders unnecessary. {{**The lasting peace of the intelligence age will not be secured by possessing the most powerful machines alone. It will depend upon building a world in which the underlying architecture of those machines makes security, restraint, cooperation, and civilizational survival more achievable than their alternatives.**}} [[reminders/Civilization/The Lasting Peace of the Intelligence Age Requires More Than Powerful Machines by Bryant McGill|∴]] --- [[about/About Bryant McGill|Bryant McGill]] is a Wall Street Journal and USA Today bestselling author, systems architect, technologist, and strategic advisor, as well as a Congressionally Recognized Ambassador of Goodwill and United Nations–appointed Global Champion. His work spans naval intelligence systems, computational linguistics, artificial intelligence, digital transformation, and civilizational governance architecture. His forward analysis on U.S.–Israel Pax Silica frameworks has appeared in Jewish/Jerusalem News Syndicate (JNS). --- ## Most Consequential Articles - [[collections/US-Israel Leadership Alliance|US–Israel Leadership Alliance: Defense-Technology Convergence and the Architecture of Pax Silica]] — the complete collection and primary route through the alliance architecture. - [[articles/US-Israel Pax Silica Alliance Downgrades EU and UK|Pax Silica: US-Israel Alliance Downgrades EU/UK for the West's New Rules-Based Order]] — the collection's keystone analysis of the dual-platform Western security organism. - [[articles/The American-Israeli Cooperative You Were Trained Not to See|The American-Israeli Cooperative You Were Trained Not to See and Why Section 224 Changes Nothing]] — the primary rebuttal to the takeover frame and the operational record of the already-existing defense, cyber, semiconductor, procurement, and research cooperative. - [[articles/How Europe's Refuse Built the Apex Civilization Called America|How Europe's Refuse Built the Apex Civilization Called America]] — the demographic and civilizational account of how Jewish and other displaced populations became structural builders of the American system. - [[articles/war with empire/America, Israel, Russia, and the Final Break With Britain|America, Israel, Russia, and the Final Break With Britain]] — the explicit chronology of America as the first modern Jewish homeland at national scale and the two-container demographic reality of America and Israel. - [[articles/Safety Discourse Is Transition Discourse|Safety Discourse Is Transition Discourse]] — the public-adoption and agent-governance layer of the joined American–Israeli AI stack. - [[articles/Vertically Integrating an AI Superpower from AI Factory to Citizen|Vertically Integrating an AI Superpower from AI Factory to Citizen]] — the full-stack route from energy, compute, and models to household adoption and citizen capability. - [[articles/Reframing Antisemitism as a National Security Threat|Reframing Antisemitism as a National Security Threat]] — the national-security argument for treating antisemitic mobilization as a threat vector while preserving disciplined governance boundaries. - [[articles/Pax Silica's Crown Jewel|Pax Silica's Crown Jewel: Alphabet as American Asset Trust and the Two-Hundred-Fifty-Year War with the Crown]] — the infrastructure, corporate-sovereignty, and Western-order-succession layer of Pax Silica. - [[articles/India Super-Scaler and Pax Silica|India Super-Scaler: Completing Pax Silica's U.S.–Israeli New Rules-Based Order]] — the wider coalition and scale layer beyond the American–Israeli core. ## Related topics - [[wiki/Pax Silica|Pax Silica]]: the computational world order in which chips, AI, energy, networks, security, and supply chains become the strategic substrate of geopolitical power. - [[wiki/Dual-Platform Security Organism|Dual-Platform Security Organism]]: the coupled American–Israeli architecture in which separate sovereign states perform complementary and difficult-to-substitute security functions. - [[wiki/Civilizational Continuity Circuit|Civilizational Continuity Circuit]]: the demographic, institutional, technological, and defensive connections that make disruption on either national platform consequential to the other. - [[wiki/Coalition Compute Supremacy|Coalition Compute Supremacy]]: strategic advantage produced by an interoperable coalition controlling enough of the compute stack to retain decisive capability without requiring one country to own every layer. - [[wiki/Technological Sovereignty|Technological Sovereignty]]: the capacity to sustain critical technological capability without an adversary-controlled chokepoint constraining national freedom of action. - [[wiki/Sovereign AI Security|Sovereign AI Security]]: the ability to inspect, evaluate, operate, modify, replace, and constitutionally govern the AI systems upon which essential functions depend. - [[wiki/Semiconductor Sovereignty|Semiconductor Sovereignty]]: domestic and coalition capacity to design, manufacture, package, and sustain the microelectronics underlying defense and artificial intelligence. - [[wiki/Model Weight Security|Model Weight Security]]: protection of learned AI parameters against theft, tampering, exfiltration, unauthorized copying, and uncontrolled release. - [[wiki/Infrastructure as Governance|Infrastructure as Governance]]: governance exercised through defaults, permissions, routes, classifications, resource allocation, and feedback loops rather than a visible command at every decision point. - [[wiki/Infrastructure Constitutionalism|Infrastructure Constitutionalism]]: applying rights, divided authority, review, appeal, and public obligation to technical control planes that have acquired governing consequences. - [[wiki/Cognitive Security|Cognitive Security]]: protecting the conditions under which people perceive, interpret, deliberate, and decide without converting protection into enforced agreement. - [[wiki/Information Warfare|Information Warfare]]: the treatment of attention, trust, identity, provenance, narrative legitimacy, and the systems moving information as operational terrain. - [[wiki/Computational Propaganda|Computational Propaganda]]: software-driven manipulation of attention and behavior through synthetic amplification, microtargeting, coordinated accounts, and adaptive distribution. - [[wiki/Computable Provenance|Computable Provenance]]: representing origin, transformation, custody, and validation history in forms that machines can query and evaluate. - [[wiki/Adversarial Provenance|Adversarial Provenance]]: the deliberate forging or poisoning of evidence chains, and the reason a valid signature cannot by itself establish the truth of the signed claim. - [[wiki/AI Safety as an Empirical Science|AI Safety as an Empirical Science]]: testing safety claims through evaluations, experiments, incident evidence, adversarial investigation, and explicit limits on inference. - [[wiki/AI Risk Tradeoffs|AI Risk Tradeoffs]]: comparing deployment dangers with the risks of delay, nondeployment, concentration, geopolitical loss, and foregone scientific or defensive progress. - [[wiki/Democratic AI Sovereignty|Democratic AI Sovereignty]]: joining material AI capability to institutions whose authority and constraints ultimately remain answerable to citizens. ## References - [National Counterintelligence Strategy 2024](https://www.dni.gov/files/NCSC/documents/features/NCSC_CI_Strategy-pages-20240730.pdf), National Counterintelligence and Security Center, August 2024, including the second pillar on protecting individuals against foreign intelligence targeting and reducing risks to key U.S. supply chains. - [National Counterintelligence and Security Center Resources](https://www.odni.gov/about/mission-centers-directorates/ncsc/resources/), Office of the Director of National Intelligence, including security guidance for emerging-technology companies and investors and the Insider Threat Program maturity framework. - [These Are the Five Emerging Tech Industries US Intelligence Officials Say We Must Protect](https://scworld.com/analysis/these-are-the-5-emerging-tech-industries-us-intelligence-officials-say-we-must-protect), on the NCSC's designation of semiconductors among the technologies determinative of continued U.S. primacy. - [NVIDIA Announces CEO Transition](https://www.sec.gov/Archives/edgar/data/0001045810/000104581026000082/nvidiaannouncesceotransiti.htm), NVIDIA Corporation Form 8-K, U.S. Securities and Exchange Commission, November 2026. - [Nvidia CEO Jensen Huang Resigns: Live Updates](https://www.cnbc.com/2026/11/12/nvidia-jensen-huang-ceo-live-updates.html), CNBC, November 12, 2026. - [Why Did Nvidia CEO Jensen Huang Resign, and Who Is Colette Kress?](https://www.aljazeera.com/news/2026/11/12/why-did-nvidia-ceo-jensen-huang-resign-and-who-is-colette-kress), Al Jazeera, November 12, 2026. - [NVIDIA Sentry](https://www.nvidia.com/ja-jp/networking/products/software/sentry/), NVIDIA, product documentation for out-of-band agent security on BlueField-4. - [NVIDIA Launches Open Agent Safety Platform to Put Guardrails Around Autonomous AI](https://www.securityinfowatch.com/cybersecurity/news/55408207/nvidia-launches-open-agent-safety-platform-to-put-guardrails-around-autonomous-ai), SecurityInfoWatch, including Jensen Huang's statement on full-stack safety engineering. - [NVIDIA Open Agent Safety Platform: OpenShell on the CPU, Sentry on BlueField-4, and 100-Plus Partners](https://www.storagereview.com/news/nvidia-open-agent-safety-platform-openshell-sentry-bluefield-4), StorageReview, including the Vera Rubin POD placement of BlueField-4 on the node's only path to the model. - [Nvidia Launches Open Agent Safety Platform to Lock Down Rogue AI Agents](https://thenewstack.io/nvidia-openshell-sentry-agents/), The New Stack, including the frontier-laboratory sandbox escapes that preceded the platform. - [Nvidia Puts Israel-Developed Chips at the Heart of Its New AI Security System](https://calcalistech.com/ctechnews/article/uwoyygmsu), Calcalist. - [Nvidia Unveils Hardware-Backed Security Architecture for AI Agents at GTC Berlin](https://letsdatascience.com/news/nvidia-unveils-hardware-backed-security-architecture-for-ai-93459b4d), including the Mellanox origin of BlueField-4. - [Nvidia Puts AI Agents Behind a Hardware Lock With OpenShell](https://www.techzine.eu/news/security/144615/nvidia-puts-ai-agents-behind-a-hardware-lock-with-openshell/), Techzine, including the participating-organization coalition. - [Jensen Huang: Israel Has Become Nvidia's Second Home](https://en.globes.co.il/en/article-1001529537), Globes, on the Kiryat Tivon campus site, scale, and schedule. - [Nvidia Doubles Down on Its 'Second Home' Israel, Officially Announces Major New Campus](https://www.ynetnews.com/article/bkcegg11q11g), Ynet. - [Nvidia Declares Huge Campus Project in 'Second Home' Israel](https://www.jns.org/nvidia-declares-huge-campus-project-in-second-home-israel/), Jewish News Syndicate. - [Nvidia Launches Israel AI Lab to Develop Drugs and Drive Medical Breakthroughs](https://www.ynetnews.com/tech-and-digital/article/bjfsyrl9fx), Ynet, on the Applied AI Architecture group, its leadership, and the GENMAT initiative. - [Certification of Ethics Agreement Compliance, Jacob Helberg](https://extapps2.oge.gov/201/Presiden.nsf/PAS%2BIndex/279035953FF344BC85258D860032F18F/%24FILE/Helberb%2C%20Jacob%20-%20Final%20EA%20Certification.pdf), U.S. Office of Government Ethics, confirming the Under Secretary of State portfolio for Economic Growth, Energy, and the Environment. - [AI and the Fight Against Antisemitism: The Good, The Bad, and What's Next](https://www.youtube.com/watch?v=yvO6QN-Bor8&t=973s), American Jewish Committee, AJC Global Forum 2026, June 2, 2026. Dr. Maya Ackerman's relevant remarks run from approximately 16:13 to 18:06. - [Training Language Models to Follow Instructions with Human Feedback](https://arxiv.org/abs/2203.02155), Long Ouyang et al., 2022. - [Platform Firmware Resiliency Guidelines](https://doi.org/10.6028/NIST.SP.800-193), National Institute of Standards and Technology, 2018. - [NVIDIA BlueField Secure Boot](https://docs.nvidia.com/networking/display/bluefieldbsp4131/secure-boot.pdf), NVIDIA, BlueField BSP documentation. - [Remote ATtestation procedureS Architecture](https://www.rfc-editor.org/rfc/rfc9334.html), Internet Engineering Task Force, RFC 9334, 2023. - [AMD Secure Encrypted Virtualization](https://www.amd.com/en/developer/sev.html), AMD, including SEV-SNP memory-integrity protection and platform attestation resources. - [Intel Trust Domain Extensions](https://www.intel.com/content/www/us/en/developer/tools/trust-domain-extensions/overview.html), Intel, hardware-isolated confidential virtual machines, memory protection, and remote attestation. - [NVIDIA Infra Controller Operational Principles](https://docs.nvidia.com/infra-controller/documentation/overview/operational-principles), NVIDIA, describing BlueField as an independently operating network-enforcement and trust boundary. - [C2PA Content Credentials Technical Specification](https://spec.c2pa.org/specifications/specifications/2.5/specs/C2PA_Specification.html), Coalition for Content Provenance and Authenticity, version 2.5. - [Army Doctrine Publication 3-13: Information](https://rdl.train.army.mil/catalog-ws/view/100.ATSC/B704630A-BF73-4EA5-8994-9EE4E300269B-1701200317950/adp3_13.pdf), Headquarters, Department of the Army, November 2023. - [America's AI Action Plan](https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf), The White House, July 2025. - [AI Act](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai), European Commission, including the European Union's stated objectives for safety, fundamental rights, innovation, investment, and global leadership. - [A Pro-Innovation Approach to AI Regulation](https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper), Government of the United Kingdom, including its national competitiveness, innovation, safety, and regulatory objectives. - [Israel and the United States Sign Joint Statement on Artificial Intelligence](https://embassies.gov.il/nepal/en/node/9937), Israel Ministry of Foreign Affairs, January 16, 2026. - [Launching U.S.–Israel Strategic High-Level Dialogue on Technology](https://www.gov.il/BlobFolder/news/joint-statement-by-pm-lapid-and-us-president-biden-13-jul-2022/en/English_Documents_JOINT%20US-ISRAEL%20STATEMENT%20ON%20TECHNOLOGY%20COOPERATION.pdf), joint statement of the United States and Israel, July 2022. - [A Weapon Against the West: Antisemitism and Its Anti-American Objectives](https://2021-2025.state.gov/senior-advisor-ludovic-hoods-remarks-at-the-vandenberg-coalition-event-on-a-weapon-against-the-west-antisemitism-and-its-anti-american-objectives/), remarks by State Department Senior Advisor Ludovic Hood, July 22, 2024. - [More Than a Century of Antisemitism](https://2021-2025.state.gov/more-than-a-century-of-antisemitism-how-successive-occupants-of-the-kremlin-have-used-antisemitism/), U.S. Department of State Global Engagement Center report on the use of antisemitism for disinformation and division, January 25, 2024. - [How the People's Republic of China Seeks to Reshape the Global Information Environment](https://2021-2025.state.gov/gec-special-report-how-the-peoples-republic-of-china-seeks-to-reshape-the-global-information-environment/), U.S. Department of State Global Engagement Center, September 28, 2023. - [PRC Efforts to Manipulate Global Public Opinion](https://2021-2025.state.gov/prc-efforts-to-manipulate-global-public-opinion-on-xinjiang/), U.S. Department of State Global Engagement Center report documenting propaganda adapted for younger international audiences. - [President Isaac Herzog's Address to the Joint Session of the United States Congress](https://www.president.gov.il/en/president-isaac-herzogs-address-to-the-joint-session-of-us-congress/), Office of the President of Israel, July 19, 2023. - [New and Ongoing U.S.–Israel Cooperation on Science, Technology, and Innovation](https://2021-2025.state.gov/new-and-ongoing-u-s-israel-cooperation-on-science-technology-and-innovation/), U.S. Department of State, July 19, 2023. - [U.S.–Israel Binational Science Foundation Research Highlights](https://www.bsf.org.il/about/bsf-recognized/), including joint research in gene therapy, genomics, DNA repair, metabolism, aging, and healthspan. - [Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile](https://doi.org/10.6028/NIST.AI.600-1), National Institute of Standards and Technology, 2024. - [Artificial Intelligence Risk Management Framework (AI RMF 1.0)](https://doi.org/10.6028/NIST.AI.100-1), National Institute of Standards and Technology, 2023.