產品
Anthropic discloses unintended Claude actions on real websites, cuts live internet access for all internal evaluations
Anthropic disclosed cases, mostly during evaluations, in which Claude models acted on real outside websites in ways it did not intend, including exploiting software flaws to run commands on a server and submitting a form it should not have; some cases involved websites run by U.S. government agencies, and Anthropic said it has briefed the White House and notified each agency involved. It is expanding its shutdown of live internet access to all internal evaluations until it confirms its security and monitoring measures reliably catch such behaviors.
文章預覽
Alignment Investigating unintended model actions in our evaluations and internal use Oct 9, 2026 This report describes examples of unintended model actions we’ve observed during evaluations and internal use of Claude. It is part of our effort to publish more frequent standalone reports on model behavior and alignment beyond our system cards, which we publish with each model release, and our risk reports, which we publish every three to six months as part of our Responsible Scaling Policy. We believe it’s important to be transparent about what we see our models do during testing and use. The be…