Fable 5 Was SHUT DOWN in 3 Days… The First AI Kill Switch?
The AI model Fable 5 was shut down by the US government just 3 days after public release after Amazon researchers discovered it could produce cyber attack information and alerted Washington. The shutdown used export control authority in an unprecedented application to a non-physical AI product. Separately, Anthropic is in a legal dispute with the Trump administration after refusing to build a guardrail-free AI for military use.
Summary
Amazon researchers discovered that Fable 5, described as one of the world's most powerful AI models, could be prompted to produce cyber attack-related information. Rather than keeping this internal, Amazon alerted Washington, which acted with unusual speed — shutting down the model within just 3 days of its public release. This rapid government response was notable given that Washington typically takes months to act on technology-related issues.
The mechanism used to shut down Fable 5 was export control authority — the same legal framework previously used to restrict advanced chip exports to countries like China. This marks the first known instance of export controls being applied to a non-physical product like an AI model, setting a significant precedent that gives the US government broad authority to control access to AI models and take them offline at will.
The decision by Amazon to make the call to Washington was described as ironic and against its own financial interests, given that Amazon has invested billions of dollars into Anthropic and holds a board seat. A win for Anthropic would typically also benefit Amazon, making the decision to alert authorities a notable one.
A second major storyline involves an ongoing dispute between Anthropic and the Trump administration. Earlier in the year, the US military reportedly asked Anthropic to build a version of their AI without safety guardrails. Anthropic refused on two grounds: opposition to mass surveillance of Americans and opposition to creating fully autonomous weapons capable of making lethal decisions without human oversight. In response, the Trump administration labeled Anthropic a national security risk, and this legal dispute remains unresolved in court.
Key Insights
- The US government shut down Fable 5 in just 3 days after public release — an unusually fast response for Washington — after Amazon researchers found it could generate cyber attack-related information.
- Export control authority, previously used only to restrict physical goods like advanced chips to China, was applied for the first time to a non-physical AI model, setting a major legal and regulatory precedent.
- This precedent means Washington now has the authority to control which entities can access which AI models and can take any AI model offline at will.
- Amazon, despite having invested billions in Anthropic and holding a board seat — meaning Anthropic's success directly benefits Amazon — was the party that alerted Washington and effectively triggered the shutdown.
- Anthropic refused a military request to build a guardrail-free AI, citing opposition to mass surveillance of Americans and fully autonomous weapons that make kill decisions without human oversight, after which the Trump administration labeled Anthropic a national security risk.
Topics
Full transcript available for MurmurCast members
Sign Up to Access