On 8 October, Anthropic launched OSS Scanner. Projects that opt in can have their public code scanned regularly by the company's strongest AI models.
A vulnerability is a flaw that may let an unauthorised person read, change or disrupt a system. When many apps rely on the same open code, one flaw can spread widely. Free scanning could therefore help small projects that lack the time or money for this work.
The service has an important limitation. Reports are produced entirely by AI and are sent without human review or triage. Anthropic says some may contain errors, such as an incorrect severity rating. Maintainers still need to reproduce the flaw, assess the risk and check a proposed fix before changing anything.
This is not antivirus software for ordinary computers, nor a guarantee of secure code. It is an extra pair of machine eyes for people who maintain open projects – useful when every alert is treated as a lead to verify, not a final answer.