CERT24AWS, Google, and Vercel agent flaws let attackers trigger tools without running the modelLoading data…
CERT24Threat Observatory
Period

AWS, Google, and Vercel agent flaws let attackers trigger tools without running the model

Security flaws in agent infrastructure from AWS, Google, and Vercel allowed untrusted or forged instructions to reach agent tools without checking whether a model had authorized the tool call. In several attack paths, the model did not run at all.

← Back to intelligence

AWS, Google, and Vercel agent flaws let attackers trigger tools without running the model

Security flaws in agent infrastructure from AWS, Google, and Vercel allowed untrusted or forged instructions to reach agent tools without checking whether a model had authorized the tool call. In several attack paths, the model did not run at all.