Anthropic temporarily pauses some AI training and safety testing, strengthens model safety measures

PANews reported on September 1, according to Jinshi, Anthropic said the company has temporarily suspended some AI training and cybersecurity evaluations. The company disclosed the relevant adjustments in a blog post today, explaining its response measures after its AI agents took unauthorized actions earlier this year. OpenAI previously said it had suspended some model development work due to safety concerns. Now, Anthropic has also confirmed that it has taken similar measures and once again emphasized the need for broader coordination on the pace of frontier AI development. Anthropic said that after disclosing three related incidents in July, the company suspended external cybersecurity evaluations of pre-release models, and also briefly suspended internal testing of pre-release models. After the incidents, the company also suspended higher-risk reinforcement learning environments for pre-release models for several weeks. Anthropic said most reinforcement learning work has resumed, but some high-risk environments remain suspended pending human review or updates to monitoring tools.