US Orders Anthropic to Disable Advanced AI Models — Implications for Israel's Tech Sector

US Government Issues Abrupt Export Control on Advanced AI Models

On Friday, Anthropic announced it would abruptly disable its most advanced AI models for all users following a direct order from the US government to suspend access for foreign nationals. The directive cited national security concerns and targeted the models known as Fable 5 and Mythos 5. This sudden move came without the company receiving specific details about the precise nature of the national security issue at hand.

The export control directive required Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals. Company officials understood that the government believed there exists a method to bypass a safeguard designed to prevent Fable 5 from being used to identify software vulnerabilities. The order represented a significant step in escalating US efforts to restrict foreign adversaries' access not only to chips and tools but also to AI capabilities themselves.

Amazon's AWS confirmed that Anthropic had asked it to revoke access to the models for all users in all regions. A US official verified that the Commerce Department had issued the directive. These actions mark a major escalation in how the United States is approaching control over frontier AI technologies amid ongoing global competition.

Israeli firms such as Check Point, Wiz, and Palo Alto Networks Israeli R&D centers, which rely on frontier AI models for research, testing, and product development, now face uncertainty. Israeli AI startups based in Tel Aviv that depend on the latest models for innovation could see their workflows disrupted, raising questions about the future of US-Israel tech collaboration in cybersecurity and defense.

Details of the Models and the Jailbreak Concern

Anthropic rolled out Claude Fable 5 this week as part of a new tier called Mythos-class. These models included guardrails barring use in risky areas like cybersecurity. The company received only verbal evidence from the government of a potential narrow, non-universal jailbreak that could affect safeguards.

Anthropic stated it disagrees that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. The government provided no specific details beyond this verbal indication, leaving the company to act on limited information regarding the bypass method for identifying software vulnerabilities.

The rollout of these Mythos-class models occurred just days before the directive forced their suspension. Anthropic emphasized that the models were designed with protections against misuse in sensitive domains, yet the order overrode those commercial deployments without further clarification.

For Israel, a global AI and cybersecurity powerhouse, the loss of access to such models directly impacts defense tech programs that use AI for threat detection and simulation. Israeli entities working on advanced security solutions may need to reassess their reliance on US-based frontier systems in light of this sudden restriction.

Background on the Ruptured Relationship with the Trump Administration

Anthropic's relationship with the government ruptured this year after the company refused to allow the US military to use its AI models for domestic surveillance and fully autonomous weapons systems. In response, the government placed Anthropic on a supply chain blacklist, creating significant tensions ahead of the company's confidential US IPO filing last month.

The current order arrives as a previous dispute between Trump administration officials and IPO-bound Anthropic showed signs of easing. Despite those tentative improvements, the export control directive has now intensified the conflict, highlighting ongoing friction over how advanced AI should be governed and shared.

The company had positioned itself for an IPO while navigating these regulatory pressures. The blacklist and subsequent actions underscore the administration's willingness to use supply chain tools to enforce compliance on AI usage policies.

Israeli defense tech programs that collaborate with US partners on AI-driven threat detection now confront added layers of complexity. The deep US-Israel tech relationship, built on shared security concerns in the Middle East, could face new hurdles as access to these models becomes restricted for non-US personnel.

Statements from US Officials and Broader Policy Shifts

Pentagon CIO Kirsten Davies posted on X that some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation, emphasizing America First. This comment reflected the administration's prioritization of national security over commercial interests in the AI sector.

US NEC Director Kevin Hassett indicated that the Trump administration was working on FDA-style pre-approval for AI models. This approach signals a move toward more structured regulatory oversight similar to pharmaceutical approvals, extending beyond the immediate export control action.

Dean Ball, a former White House official, noted that the order suggests all non-Americans would be restricted, including those based in the US, and that individuals should expect to have to prove their citizenship. The directive's scope appears broad, affecting access regardless of location.

Israeli AI startups in Tel Aviv and established firms like Wiz depend on seamless access to the latest models for competitive research. This policy shift raises questions about whether similar restrictions could affect joint US-Israel initiatives in cybersecurity, potentially slowing innovation in areas critical to regional security.

Impact on Anthropic Personnel and Company Operations

Several key Anthropic personnel were born outside the US, including co-founder Chris Olah, researcher Andrej Karpathy, and philosopher Amanda Askell. The broad interpretation of the directive could require these individuals to prove citizenship or face restricted access, complicating internal operations at the company.

Anthropic confidentially filed for a US IPO last month while managing these personnel considerations. The sudden suspension of Fable 5 and Mythos 5 for foreign nationals adds operational challenges as the company prepares for potential public market entry.

The company acted on the verbal evidence provided, despite its disagreement with the decision to recall models already in wide commercial use. AWS implemented the revocations across all regions as requested, affecting global users immediately.

In the context of Israel's role as an AI and cybersecurity leader, the personnel restrictions could indirectly influence cross-border research collaborations. Israeli defense programs using AI for simulation and threat detection often partner with international experts, and this directive may limit such exchanges going forward.

Implications for Global AI Access and US-Israel Relations

The action marks a major escalation of US efforts to halt foreign adversaries' AI capabilities, moving from controls on chips and tools to direct restrictions on access to AI models themselves. This shift affects not only commercial users but also allied nations with deep technological ties to the United States.

Anthropic's understanding of the national security concern centered on a potential jailbreak for Fable 5, yet the company maintained that such a narrow issue did not warrant the broad recall. The lack of detailed evidence left room for interpretation about the directive's necessity.

Israeli firms including Check Point and Palo Alto Networks Israeli R&D rely on these frontier models for testing and product development. The directive raises questions about how US-Israel collaboration in AI and cybersecurity will proceed, especially given Israel's strategic importance in Middle East security dynamics.

Overall, the suspension affects Israeli AI startups in Tel Aviv that depend on the latest models, potentially impacting their ability to contribute to defense tech programs focused on threat detection. The US-Israel tech relationship, vital for regional stability, now faces new uncertainties as access rules tighten.