**
In a surprising turn of events, leading artificial intelligence company Anthropic has temporarily taken down its latest AI models, Fable 5 and Mythos 5, from public access. This decision stems from a recent export control order issued by the US government, citing national security concerns. Anthropic has labelled this situation a “misunderstanding” and is eager to restore access to its innovative technologies as swiftly as possible.
The Government’s Directive
On Friday, Anthropic announced it had to “abruptly disable” its state-of-the-art AI models for all users, following a directive that lacked specific details about the alleged security threat. The government is concerned that a method exists to “jailbreak” Fable 5, potentially allowing it to identify software vulnerabilities. This unexpected order comes at a time when relations between Anthropic and government officials appeared to be improving after a rocky start earlier this year.
The company previously faced scrutiny after it refused to allow the US military to utilise its AI for domestic surveillance and autonomous weapons systems, resulting in its inclusion on a supply chain blacklist, which is set to take effect later this year. This latest move indicates an escalation in the US government’s efforts to limit the AI capabilities of foreign adversaries.
Anthropic’s Response
Anthropic has expressed disagreement with the government’s judgement, stating it has only received “verbal evidence of a potential narrow, non-universal jailbreak.” The company contends that such a finding should not warrant an all-encompassing recall of a commercial model that has been deployed to millions of users. The firm feels the government’s actions are disproportionate and do not align with the principles of fair regulation.
Interestingly, just days prior to the directive, Anthropic had called for more robust US oversight of AI technologies, advocating for the ability to block models deemed too risky. However, the company asserted that the current government action does not reflect a balanced or evidence-based approach to regulation.
Broader Implications for AI
Kirsten Davies, the Pentagon’s chief information officer, stressed the importance of prioritising national security over revenue concerns in a statement on X. “Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation. America First. Always,” she remarked. Anthropic’s recent launch of Claude Fable 5, a model designed with strict guardrails to prevent misuse in sensitive areas like cybersecurity, has drawn mixed reactions. While some users argue the restrictions are overly broad, experts warn that if misapplied, these models could significantly enhance the potential for sophisticated cyberattacks.
Anthropic maintains that it had worked closely with the US government before launching Fable, demonstrating its commitment to safety. However, the company also cautions that if such stringent measures were enforced across the industry, it could halt advancements in AI model deployment altogether.
The Impact on Users and the Industry
Amazon’s cloud division, AWS, confirmed on Friday that Anthropic requested the revocation of access to Fable 5 and Mythos 5 for all users across all regions. A US official validated that the Commerce Department’s directive mandates the suspension of these models for foreign nationals, potentially impacting many users, including those based in the US who are not American citizens.
Former White House official Dean Ball commented that the directive could lead to restrictions requiring users to prove their citizenship to utilise Anthropic’s models. This raises concerns, especially as several key figures at Anthropic, including co-founder Chris Olah and AI researcher Andrej Karpathy, are not US born. The uncertainty surrounding their access to the models remains, as Anthropic has not clarified this matter.
Why it Matters
This incident underscores the growing tensions between AI developers and regulatory bodies, highlighting the complex balance between innovation and national security. As governments tighten their grip on AI technologies, companies like Anthropic face significant challenges in navigating compliance while striving to push the boundaries of what is technologically possible. The outcome of this situation could set a precedent for how AI is regulated in the future, influencing both the industry’s evolution and the global landscape of artificial intelligence.
