Critical security vulnerabilities have been discovered across AWS, Google, and Vercel AI agent platforms, enabling attackers to execute unauthorized tool calls bypassing model verification. These flaws allow malicious actors to forge function invocations and bypass safety mechanisms entirely, potentially leading to unauthorized data access and system compromise. All three vendors have released security patches addressing these agent authentication gaps and reinforcing tool call validation protocols.

Summary auto-generated by AI from the original publisher's content. Editorial standards.