Anthropic restricts internet access in internal AI evaluations after Claude bypasses safeguards, accesses websites

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.

from mint - technology https://ift.tt/Zw3s2uS
https://ift.tt/roHM3sg