Anthropic’s Mythos AI Model Is Being Accessed by Unauthorized Users - Bloomberg
Security researchers found that Anthropic’s new Mythos AI model was reachable by unauthorized users through exposed API endpoints. This lapse could expose sensitive prompts and responses, prompting Anthropic to investigate and strengthen its access controls.
- Anthropic's Mythos AI model was accessed by unauthorized users after API keys leaked onto public Slack channels
- Anthropic rotated all impacted keys, shut down mismatched sessions, and tightened authentication after detecting unusual traffic
- The company hasn't disclosed how many keys leaked or how many unauthorized calls were made
- Some enterprise customers paused rollouts pending clearer safeguards on key security
Anthropic’s new Mythos model slipped into the hands of people outside its approved user base after some access keys landed on public Slack channels. Security researchers and hobbyists picked up on the leaked credentials and started sending prompts to the model. That meant code execution, text generation and other capabilities intended only for paying customers wound up in wide circulation. Anthropic noticed unusual traffic patterns and confirmed the breach late last month.
The company moved fast to rotate every impacted API key and shut down sessions that didn’t match its customer lists. Behind the scenes, engineers also tightened the authentication flow and flagged any new keys that suddenly started firing off hundreds of requests. So far, Anthropic hasn’t said how many keys leaked or how many calls unauthorized users made, but it’s clear the misstep exposed some of Mythos’s most advanced features.
Meanwhile, users in tech forums have posted examples of Mythos generating code snippets, writing marketing copy and even drafting legal clauses. Those samples revealed its raw capability—sometimes impressive, other times prone to typical AI quirks like making up facts. Anthropic now plans a formal audit of its internal key-management processes and will require multi-factor checks before issuing new tokens.
Investors and enterprise customers have taken note. Some corporate buyers paused their rollouts until they get a clearer picture of how Anthropic will keep secret keys under tighter lock and key. Internally, the company is weighing a bounty program for anyone who spots similar leaks in the future.