Blog

According to Wired, the administration’s position is supported by findings from the National Security Agency (NSA). The agency reportedly concluded that skilled users could disable safeguards built into Fable 5, safeguards specifically designed to prevent access to certain advanced capabilities inherited from Anthropic’s more powerful Mythos model. Officials fear that if those protections are bypassed, the technology could potentially be used in areas such as advanced hacking, cybersecurity exploitation, chemistry research or biological experimentation. Rather than attempting to identify every possible vulnerability themselves, government agencies reportedly expect Anthropic to proactively strengthen the model’s security systems.