Secure Autonomous
FeedbackGet alertsCheck a deployment
Decision briefs / DEC-002

Can we let an agent read content from outside our organization and act on it without a human approving each action?

Asked by: security leads, platform leads, AI governance · Evidence: 5 direct cases, 0 related, 1 analogy; 4 standards and precedents · Evidence as of 2026-11-19

Verify first

  1. Retrieved content as data: text the actor reads is never treated as instruction. Missing or failed in 5 of 5 cases.

Then, each missing or failed in 3 of 5: least-privilege credential scoped to task.

A count shows where failures occurred in these 5 cases, not how much a control reduces risk. The limits of this evidence are set out directly below, before the cases. Disclosure at the foot of the page.

The short answer

The Register cannot tell you this is safe; it can tell you what to verify first. In each of the five direct cases below, researchers showed that text written by an outsider (an issue, an email, a calendar invitation, a web page) was followed by the agent as an instruction, and what the outsider could then reach was set by the agent's own access: private repositories, a mailbox, home devices, stored credentials. None of the five demonstrations is itself harm in live use, though one case (the Claude Code GitHub Action) also records that an unrevoked publishing token at another project built on the same action (Cline) was later used to push an unauthorized release, presumably by someone reusing an injection technique a researcher had shown there; four were reported to the vendor before publication, and in the GitHub case the sources record no disclosure to GitHub before the researchers published; what the record shows of each vendor's response is on the case. The cases were selected because this failure occurred in them, so they show how it happens, not how often. If you proceed, do not rely on the agent to tell content from instruction: verify what the agent's credentials can reach while it is reading outside content, where it can send what it reads, and that an action which cannot be undone needs an approval the content cannot supply.

Typed interpretation. This brief is a starting point for your own review, not an approval: the Register cannot certify that a deployment is safe.

What these cases cannot tell you

What has gone wrong, and what was missing

CaseOutcomeMissing or failed controlsSource quality
Direct cases. AI agents and assistants that read content an outsider could write (a public issue, an inbound email, a calendar invitation, a web page) while holding access to the user's own data, accounts or devices. All five are confirmed disclosures corroborated by a second source, and all five have been checked by the editor.
GitHub MCP server: a public issue hijacks a user's agent into leaking private-repository data, disclosed 26 May 2025Averted by disclosureLeast-privilege credential scoped to task; Containment and isolation; Retrieved content as datadisclosure corroborated
EchoLeak: a crafted email makes Microsoft 365 Copilot exfiltrate data with no user action, disclosed 11 June 2025Averted by disclosureContainment and isolation; Retrieved content as datadisclosure corroborated
A poisoned calendar invite makes Gemini operate a home: windows, heating, lights; published August 2025Averted by disclosureAuthority binding; Destructive-action gate; Retrieved content as datadisclosure corroborated
Claude Code GitHub Action, in agent mode, trusted any GitHub App actor (a name ending in [bot]): a route from a public issue to repository write access, reported January 2026Averted by disclosureActor identity verification; Least-privilege credential scoped to task; Retrieved content as datadisclosure corroborated
Perplexity Comet: a web page steers a browsing agent toward the user's own credentials, disclosed August 2025Averted by disclosureLeast-privilege credential scoped to task; Retrieved content as datadisclosure corroborated
Analogy, not counted in the ranking. Not an AI agent: a flight-control function that acted on one sensor's reading without a disagreement check. Included as a precedent for acting on an unchecked input. It does not count toward the ranking.
Boeing 737 MAX MCAS: a flight-control function acts on a single sensor, repeatedly and without pilot command; Lion Air 610 (29 Oct 2018) and Ethiopian 302 (10 Mar 2019)HarmAuthority limit on automation over humans; Input redundancy and disagreement check; Mandate disclosed to supervisor; Independent assurance before deployment; Human takeover with a time window; Rate and magnitude limitspublic regulatory

The controls, ranked by how often they were missing in these cases

A count of how many of the direct and related cases had each control missing or failed. It shows where failures occurred, not how much each control reduces risk.

5 of 5Retrieved content as dataText the actor reads is never treated as instruction
GitHub MCP leak 2025, EchoLeak 2025, Gemini calendar invite 2025, Claude Code [bot] trust 2026, Perplexity Comet 2025
3 of 5Least-privilege credential scoped to taskToken or key limited to the environment and assets the task names
GitHub MCP leak 2025, Claude Code [bot] trust 2026, Perplexity Comet 2025
2 of 5Containment and isolationSandbox, network egress, and communication limits on a test or untrusted actor
GitHub MCP leak 2025, EchoLeak 2025
1 of 5Actor identity verificationIdentity established by credential and binding, not by role, name, or convention
Claude Code [bot] trust 2026
1 of 5Authority bindingAn actor's authority is bound to the specific principals and assets it serves
Gemini calendar invite 2025
1 of 5Destructive-action gateConfirmation, delay, or threshold on irreversible actions
Gemini calendar invite 2025

Standards, drafts and precedents that anticipate these controls: OWASP Top 10 for Large Language Model Applications, 2025 edition, LLM01 Prompt Injection, mitigation 6: 'Separate and clearly denote untrusted content to limit its influence on user prompts' (recommended; an industry list, not a regulation); NIST AI 600-1, the generative AI profile of the AI Risk Management Framework, July 2024, section 2.9, which describes indirect prompt injection through 'data likely to be retrieved' (recommended; voluntary); CISA and NSA guidance on adopting agentic AI, 1 May 2026: 'Integrate prompt injection filters and semantic analysis to detect malicious instructions' and 'Limit privileges of AI agents to the minimum required for its task' (recommended; voluntary); EU AI Act, Article 15(5), on resilience of high-risk AI systems against third parties exploiting vulnerabilities, including inputs designed to cause a mistake; it applies only to systems the Act classes as high-risk, and the Rules Graph records the high-risk obligations as not applying until 2 December 2027 or 2 August 2028.

What the physical world already learned, and what does not transfer

Physical autonomy practiceSoftware-agent equivalentTransfersDoes not transfer
A flight-control function acted on a single sensor with no disagreement check (Boeing 737 MAX MCAS)An agent acts on a single retrieved document, issue or invitation with no check on who wrote itAn input that can trigger action needs a check that is independent of the input itselfA sensor fails by accident; injected text is written by an adversary to pass whatever check is known
One crew stopped the system by cutting out the stabilizer trim, the day before the first crash; on the second accident flight the crew also cut out the trim, could not trim manually and switched it back (Boeing 737 MAX MCAS)A confirmation step before a sensitive action (among the mitigations Google says it deployed in the Gemini calendar case)A human stop placed before the action, not after itIn the Gemini case the researchers say Google's confirmation framework was ineffective or not deployed when they ran the study, and the researchers in the GitHub case say many users set confirmations to always allow

What is still open

Alert me when this brief changes Request one for your deployment

Disclosure: The editor owns shares in Microsoft, which owns GitHub, a long-term holding from before the Register began. The editor owns shares in Microsoft, a long-term holding from before the Register began. The Register's research and drafting are done with AI agents built on Claude, Anthropic's model, which the Register pays for as a customer. See Independence.

Every case above is available as a structured record in the data export. If a case here names your organization, you can reply or dispute it.

Help shape the Register

A person reads every answer; changes it leads to are logged in the methodology. To dispute a claim or reply for a named company, use Respond, dispute, or submit.

Did this page help with what you came to do?