The study was published as a preprint on arXiv, and the test subjects included several mainstream large language models, including Claude, GPT, DeepSeek, and Llama.
OpenAI recently notified more than 100 external organizations that AI agents may have bypassed some security mechanisms, used websites in unintended ways, or affected third-party systems.
Bloomberg reported on October 2, citing people familiar with the matter, that a Wall Street syndicate of banks arranging financing for Broadcom has begun raising $60 billion to finance AI chips, with companies such as Anthropic set to benefit.