AI-Generated Security Patches Found Unreliable, May Introduce New Vulnerabilities

Research has found that more than half of AI-generated security patches fail to fully fix vulnerabilities and may introduce new flaws. The study highlights risks associated with using AI models, including Anthropic’s Claude and OpenAI’s GPT-5, for automated patch management in cybersecurity. No specific dates, CVE IDs, or technical methodologies were disclosed in the findings. The impact includes potential exploitation of newly introduced vulnerabilities due to flawed AI-generated code corrections. The research underscores the unreliability of AI-driven patching in its current state.