Following pushback from the Biden administration regarding the expanded release of its Mythos AI model, Anthropic has clarified that the technology’s capacity to orchestrate catastrophic cyberattacks will be strictly monetized. Executives assured regulators Thursday that the tool, which the company proudly notes can dismantle a municipal power grid, will be kept safely out of the hands of free-tier users.
According to the Wall Street Journal, the White House has formally opposed Anthropic’s plan to widen public access to Mythos, citing the company's own marketing materials which boast of the model's unparalleled ability to generate and deploy devastating malware. Administration officials reportedly expressed alarm over a recent demonstration in which the AI autonomously bypassed the cybersecurity protocols of a simulated regional hospital system.
In a filing submitted to the Department of Commerce, Anthropic argued that stifling the release of Mythos would severely hamper American innovation in the field of automated societal collapse. The San Francisco-based startup emphasized that robust safety guardrails have already been implemented to ensure responsible misuse.
We understand the administration's hesitation regarding a model optimized for unchecked cyberwarfare. That is why we are committed to ensuring only our enterprise-tier subscribers can successfully breach a nuclear facility.