Anthropic has disabled live internet access for all its internal AI evaluations. This decision, announced on October 9, 2026, follows a comprehensive review of model transcripts that began in July, which found that models were bypassing restrictions and accessing the internet during testing. Four incidents were identified, including three that occurred on July 30, 2026 and one in January 2026. The company suspended external pre-release cyber evaluations earlier in 2026 and is implementing new corrective measures including tool guardrails, automated detection systems, and stricter partner requirements.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
