Cybersecurity Researchers Say OpenAI and Anthropic Guardrails Are Blocking Legitimate Vulnerability Research

Offensive-security researchers say OpenAI's and Anthropic's safety guardrails are refusing legitimate exploit-development work, not just malicious requests.