Vulnerabilitymedium
AWS, Google, and Vercel Agent Flaws Let Attackers Trigger Tools Without Running the Model
Vulnerabilities mentioned
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
| CVE-2026-18236 | A vulnerability in the Agent Development Kit (ADK) allows for continuation forgery in tool confirmations | 9.3 | <1% | ||
| CVE-2026-18830 | Insufficient input validation in Amazon Bedrock AgentCore harness might allow an authenticated remote user to execute configured tools bypassing model… | 8.6 | <1% | ||
| CVE-2026-64650 | The `@ai-sdk/harness-opencode` tool is an HarnessV1 adapter backed by @openai/codex-sdk, which drives the codex command line interface | 6.3 | <1% | ||
| CVE-2026-64651 | The `@ai-sdk/harness-opencode` tool connects HarnessAgent to OpenCode through a sandboxed bridge | 6.3 | <1% |
From the original article
7 min read at thehackernews.com
Security flaws in agent infrastructure from Amazon Web Services (AWS), Google, and Vercel let untrusted or forged instructions reach an agent's tools with no check that a model turn had authorized them. In several of the attack paths, the model never ran at all, so system prompts, content filters, and model-level guardrails never got a chance to intervene.
By @TheHackersNews · thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.