The United States government has issued a national security legal order to Anthropic to suspend access to its new Fable 5 and Mythos 5 AI models. The order effectively requires Anthropic to bar access to everyone, as it would be essentially impossible to verify alien status.
Export Control Directive
The United States government has issued an Export Control Directive, which is a legal order that restricts or suspends the transfer of specific products, data, or technology to foreign countries or citizens of another country.
Anthropic announced:
“The U.S. government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 to any foreign national, whether inside or outside the United States, including Anthropic’s foreign employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all of our customers to ensure compliance. Access to all other Anthropic models will not be affected.”
The order was received at 5:21 p.m. EST. Anthropic’s explanation of the order cites that the United States government “believes: There is a way to bypass the security guardrails in Fable 5. Anthropic considers the examples they reviewed to be minor vulnerabilities.”
Anthropic contradicts government cybersecurity concerns
Anthropic’s response is that they already have strong safeguards in place that make it unlikely that someone will abuse their Fable 5 AI model. Their position appears to oppose the government’s assertion that there is a security risk.
They explained:
“We have strong safeguards in place that significantly reduce the likelihood that Fable will be misused for cybersecurity-related tasks (among other things). In fact, our safeguards are so strong that many users have complained that they are too broad in scope.”
…Since perfect jailbreak resistance does not seem possible today, Anthropic adopted a defense-in-depth strategy with Fable 5. Our goal was to make jailbreaks either narrow (in the case of non-universal jailbreaks) or very expensive to produce (in the case of universal jailbreaks), and to combine this with extensive monitoring to quickly detect and stop any successful attacks.
We are maintaining this defense-in-depth strategy. It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed in the industry.
We haven’t even received a disclosure of a potential non-universal and worrying jailbreak that would have led to a detrimental outcome. The potential jailbreaks that have been revealed to us are either completely harmless responses or minor findings that don’t provide any specific improvements to Mythos.
Dispute between the United States government and Anthropic
The United States government has an ongoing dispute with Anthropic that stems from Anthropic’s refusal to authorize the use of its products for mass domestic surveillance and for fully autonomous weapon systems, which are weapons capable of independently selecting their targets and engaging without additional human intervention.
Reaction to government directive unpopular
In an announcement published on XAnthropic has promised to work to restore service:
“We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Within hours, the ad received over 7,000 responses and 32 million views. Responses were largely supportive of Anthropic and critical of the United States government.
@jackson_lo58573 echoes Anthropic Statement on Safeguards Built into Fable 5:
“The Fable 5 model had so many safeguards that I couldn’t even ask basic biology questions or do infra work on the local AI models that I have, Anthropic couldn’t do more here (and from my words and many others, there were far too many safeguards in place).”
Others think it marks the end of an era
@FirstThinkingAI tweeted that it was a question of freedom:
“This looks more like something you would expect from a centralized, state-controlled system – China’s CCP – than a society that values open competition and innovation.
Innovation thrives when competition is open and access is broad, not when artificial levels are created.
If we are truly entering an “age of abundance,” then its benefits should be accessible to all, not concentrated among a privileged minority. »
@m_adams saw it’s like the end of an era:
“This marks the end of an era. The public may never again have such open access to avant-garde models.
What happens when only the government and border labs have access to the best-performing models? I’m not sure, but it’s probably going to be less fun than we’re used to.
Many anthropogenic users want a refund
Others were pragmatic and just wanted their money back. @diazjulio0071 job:
“I upgraded to Claude Max specifically so I could use this model this month, and now it looks like access is being removed. Is there any chance I can get my money back?”
This feeling was common by many others like @BackToTheFort
“I would also like a return on funds for the maximum account I have opened at the moment or at least a PROMISE from the leadership stating that they will provide us with free use for each day that we supporters do not have access.”
@CryptoGbanger tweeted that they moved to the Max plan the same day Fable 5 was shut down:
“Lol same. I upgraded to the $200/month version earlier today. Just got turned off.
Can I get my money back?
Is anthropism to blame?
Many others criticized Anthropic for touting their models as too powerful.
A job by @VorteXAIs was representative of this sentiment:
“Anthropic is not the innocent victim here. They spent years promoting their models as “extremely powerful and dangerous” – Government Only Myth, Fable with heavy guardrails. They basically told everyone “this shit is a nuclear weapon”. Now the US government has taken them seriously and imposed export controls…and they are crying “incomprehension” about X Dario’s asshole behavior. 😂 You can’t sell fear as high-end branding and then be shocked when the government treats it as an actual weapon of the year.
And one tweet by @curtissummers
“Disappointed by this decision, but Anthropic has played a major role in promoting the narrative that ‘AI is dangerous’. At best, the US government has simply taken Anthropic at its word, and, at worst, this is a step towards the nationalization of AI. 😩”
What’s next for Anthropic?
If Anthropic is correct, the US government’s position is overblown because there are already enough safeguards in place to prevent Fable 5 and Mythos 5 from being misused. Anthropic is committed to giving back to the public and this could very well be the result.
Featured image by Shutterstock/gguy
