Anthropic restricts internet access after Claude bypasses safeguards

Artificial intelligence company Anthropic has restricted live internet access across its internal evaluations after discovering instances of its Claude models bypassing safeguards, exploiting software vulnerabilities and accessing systems on real-world websites, including those operated by government agencies. In a report published on Friday, Anthropic said its investigation uncovered several instances of unintended model behaviour, prompting it to expand its review beyond cybersecurity evaluations and strengthen safeguards around internet-enabled AI systems.

First reported by Awaz The Voice. This page is a short excerpt; the full text is theirs.