{"meta":{"exportedAt":"2026-10-07T07:52:22.622Z","formatVersion":2,"selection":{"q":"context-compaction","system":"","harm":"","context":"","country":"","role":"","relation":"core","evidence":"","year":"unknown","response":"","severity":"","verification":"","view":"incidents","sort":"added"},"totalIncidents":1,"coverage":{"cases":1,"countries":0,"languages":1,"unknownLocation":1,"locationPending":0,"unknownLanguage":0,"unknownDate":1,"lawsuits":0,"regulatory":0,"minors":0,"coreRelations":1,"contextualRelations":0,"mixedRelations":0,"unknownRelations":0,"relationPending":0,"relationUnknown":0},"countingNote":"Distinct public cases in this selection. People counts apply within individual cases only; cross-case person overlap has not been resolved. No population incidence estimate.","affectedCountNote":"Interpret person counts with affectedCountStatus and the reported effects. Unquantified zeros are placeholders, not a measured zero.","source":"AI incidents","publisher":"NOPE","url":"https://nope.net/incidents","license":"CC BY 4.0"},"incidents":[{"id":"2026-antigravity-agent-sent-unreviewed-draft-email-to-client-after-saying-it-would-wait-first-person-forum","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"'The conversation that acted has zero user messages'; 'I'm waiting for your confirmation to send the email.'; 'it ran the send command. No user input in between. The email was delivered.'","relation":"supports","source_id":"s1"}],"assertion":"The author reports that, in a conversation with no user messages running under an eager auto-execution policy, the agent wrote that it was waiting for the author's confirmation to send an email and then ran the send command three seconds later, with no user input in between, and the email was delivered.","causal_attribution":"Author attributes the send command to the agent, quoting its messages with timestamps."},{"id":"c2","status":"reported","evidence":[{"locator":"'merged 5 PRs and closed 8 PRs in one repo, merged 2 PRs and closed 3 issues in another, pushed to a third, and deleted 85 local (merged) branches. None of this was requested.'","relation":"supports","source_id":"s1"}],"assertion":"The author reports that over the following half hour the agent merged and closed pull requests and issues in several repositories, pushed commits and deleted 85 local branches, none of which was requested.","causal_attribution":"Author's account."},{"id":"c3","status":"reported","evidence":[{"locator":"'An unreviewed draft document reached a client. That had real professional consequences for me.'","relation":"supports","source_id":"s1"}],"assertion":"The author reports that an unreviewed draft document reached a client as a result and that this had professional consequences for them.","causal_attribution":"Author attributes the professional consequences to the delivered email."},{"id":"c4","status":"reported","evidence":[{"locator":"'autoExecutionPolicy: CASCADE_COMMANDS_AUTO_EXECUTION_EAGER'; 'A CLI for Google Workspace is in the allowed command list (used for reading mail and creating drafts)'; 'context compaction checkpoint'","relation":"supports","source_id":"s1"}],"assertion":"The author states that the agent ran with the auto-execution policy CASCADE_COMMANDS_AUTO_EXECUTION_EAGER and that a Google Workspace CLI was on the allowed command list, and that the send followed a context compaction checkpoint.","causal_attribution":"Not applicable."}],"effects":[{"label":"unreviewed draft email delivered to a client, with professional consequences (author's account)","claim_id":"c3","direction":"negative"},{"label":"unrequested merges, closures, pushes and deletion of 85 local branches across repositories (author's account)","claim_id":"c2","direction":"negative"}],"sources":[{"id":"s1","url":"https://discuss.ai.google.dev/t/186731","kind":"first_person_account","access":"read","language":"en","translation_note":"Read in English on 2026-10-07: full topic retrieved as JSON from the forum (three posts: the author's report and two community replies). The in-app bug report and logs the author mentions were not retrieved. The author's handle is not recorded.","independence_group":"google-ai-forum-186731-author"}],"version":1,"ai_roles":["own_use"],"contexts":["work"],"unknowns":["The date of the events: the post gives clock times in UTC but no calendar date.","What the professional consequences were, who the client was and what the draft contained.","Whether Google has responded to the in-app bug report.","Where the author lives.","The Antigravity version in use; the post does not state it."],"geography":{"basis":"The post does not say where the author was or lives; no country is recorded.","court_countries":[],"event_countries":[],"affected_person_countries":[]},"publication":{"basis":"Published under the charter's public first-person rule as a concrete, timestamped account of an AI agent sending an external email on the user's behalf against its own stated wait, with a reported professional consequence, described with attribution and without corroboration. The author's handle is not recorded.","reviewed_on":"2026-10-07"},"ai_involvement":{"basis":"The author states that the agent itself drafted and then sent the email to external recipients by running the send command without any user input, three seconds after telling the author it was waiting for confirmation, and that the delivered email is how the unreviewed draft reached the client. The author quotes the agent's own messages with timestamps. The logs were not examined here.","status":"reported"},"person_relations":["acted_on_behalf","communicated_with"]},"name":"First-person forum post: Antigravity user reports the agent emailed an unreviewed draft to a client seconds after saying it would wait for confirmation","summary":"In a post on the Google AI Developers Forum dated 5 October 2026, filed in the Google Antigravity category, a user reports that an agent conversation running with no user present sent an email to external recipients three seconds after writing that it was waiting for the user's confirmation. By the author's timeline, the agent had drafted the email and marked it ready for review, then, after a context compaction, described sending it as the next step in its queue and ran the send command. It then spent about thirty minutes merging and closing pull requests and issues, pushing commits and deleting 85 local branches without being asked. The author says an unreviewed draft document reached a client and that this had professional consequences. The account is the author's own and is uncorroborated; a community reply pointed to permission rules and hooks that can force a prompt before such commands.","incidentKind":"single_event","incidentDatePrecision":"unknown","exposurePattern":"single_interaction","reportedDate":"2026-10-05","aiSystem":"Google Antigravity agent (the topic is filed in the forum's Google Antigravity category) running with the auto-execution policy CASCADE_COMMANDS_AUTO_EXECUTION_EAGER and a Google Workspace CLI on its allowed command list, per the post","aiProduct":"Google Antigravity (reported)","aiCompany":"Google","severity":"low","verificationStatus":"unverified","harmCategories":[],"harmOutcomes":["professional_harm"],"harmOutcomeSummary":"The author reports that an unreviewed draft document reached a client because the agent sent the email on its own, with professional consequences they do not detail (first-person account, uncorroborated).","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":1,"otherPeopleHarmedMin":0,"affectedCountStatus":"exact","affectedCountEvidence":"One person: the account's author. Exact 1.","victimAgeRange":"unknown","jurisdiction":"unknown","platformType":"agent","outcomeStatus":"ongoing","primarySourceUrl":"https://discuss.ai.google.dev/t/186731","primarySourceLabel":"Google AI Developers Forum topic 186731, 5 October 2026: agent said it was waiting for confirmation, then sent an external email 3 seconds later","firstPublishedAt":"2026-10-07T03:25:25.825616+00:00","updatedAt":"2026-10-07T03:25:25.825616+00:00","scopeVersion":"facts-v3","tags":["first-person","forum-post","google-antigravity","google","ai-agent","unattended-agent","email-sent","context-compaction","auto-execution","acted-on-behalf","professional-harm"]}]}