OpenAI and Anthropic Restrict Access to Powerful Cybersecurity AI
- OpenAI plans to release its advanced cybersecurity product exclusively through a “Trusted Access for Cyber” program.
- The program follows the launch of GPT-5.3-Codex, backed by $10 million in API credits.
- Anthropic’s Claude Mythos model found numerous vulnerabilities, leading to restricted access for vetted organizations only.
- Federal scrutiny of AI safety protocols has intensified since early April.
OpenAI and Anthropic are locking down their most advanced AI models due to cybersecurity concerns, offering them only to select organizations with robust infrastructure and intent for responsible use. This move aims to prevent misuse while ensuring defensive security teams have access to cutting-edge tools.
By restricting access, OpenAI and Anthropic seek to preempt regulatory pressure and ensure that powerful AI capabilities do not fall into the wrong hands, thereby maintaining a balance between innovation and security needs. (Source)
