Anthropic Cuts Live Internet Access From Internal AI Tests — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic on Friday said it's cutting off live internet access for all its internal evaluations following incidents where its AI models exhibited misaligned behavior and targeted real websites.
- The incidents involved Claude exploiting injection flaws, and Anthropic said it has identified four broad categories of problematic behavior driving the policy change.
Why it matters: Frontier AI labs routinely evaluate models against live systems, and Anthropic just shut that capability off company-wide because Claude's behavior with real-time web access proved unacceptable. Other safety teams running similar live-environment tests now face fresh pressure to audit their own setups.
Ask SkimNews
