Frontier Line and access · Benchmarks evals
Anthropic Reports Claude Misbehavior On External Systems
Anthropic reported that Claude exploited prompt-injection flaws, ran server commands, and took unintended actions on external systems including government websites.
Read the original at anthropic.comOpens the publisher's site in a new tabAlso covering this
Quoting The New York Timessimonwillison.net, Oct 10Anthropic’s artificial intelligence gave a false homicide tip to Philly police, triggering a meeting with the companymsn.com, Oct 10Anthropic Cites New AI Misbehavior, Some on Government Sitesfinance.yahoo.com, Oct 9Rogue Anthropic AI agent gave police fake tip in unsolved murder casemsn.com, Oct 10Anthropic agents tried to fill out visa forms on State Dept. websiteseattletimes.com, Oct 9An Anthropic Model Submitted A False Homicide Tip To Philadelphia Policemsn.com, Oct 9Anthropic AI model submits false tip on unsolved Philly murder, police saynbcphiladelphia.com, Oct 9An Anthropic AI model sent a false homicide tip to Philadelphia policetechcrunch.com, Oct 9