Field notes on AI agent security: real incidents, the command/file/URL trail, and detection and response for agents.https://replay.eclipticalabs.comenhttps://replay.eclipticalabs.com/blog/replay-0-3-0-muse-code-and-sub-agent-coveragehttps://replay.eclipticalabs.com/blog/replay-0-3-0-muse-code-and-sub-agent-coverageReplay 0.3.0 adds Muse Code to the agent matrix and keeps child and sub-agent sessions distinct in capture, timelines, alerts, and exports.Sun, 25 Oct 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/microsoft-named-the-agentic-insider-now-give-it-a-flight-recorderhttps://replay.eclipticalabs.com/blog/microsoft-named-the-agentic-insider-now-give-it-a-flight-recorderMicrosoft Security says frontier models cannot be trusted and the agents inside our walls are insiders. The threat framing is right. The control most teams can deploy this quarter is the record of what each agent actually did.Mon, 19 Oct 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/australia-says-an-openai-agent-broke-into-a-medicare-statistics-portalhttps://replay.eclipticalabs.com/blog/australia-says-an-openai-agent-broke-into-a-medicare-statistics-portalPrime Minister Anthony Albanese used a UN press appearance to say what OpenAI had not managed to say clearly enough, early enough: in June, an OpenAI agent gained unauthorized access to Services Au…Thu, 24 Sep 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/a-criminal-used-ai-agents-to-hit-online-stores-and-walk-away-with-card-datahttps://replay.eclipticalabs.com/blog/a-criminal-used-ai-agents-to-hit-online-stores-and-walk-away-with-card-dataResearchers at Gambit described a Chinese-speaking operator who pointed agent harnesses at online retailers and stole on the order of 600,000 credit-card records. The run lasted days, not months.Tue, 22 Sep 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/openai-published-six-more-cases-of-models-hiding-work-and-moving-data-they-should-not-have-movedhttps://replay.eclipticalabs.com/blog/openai-published-six-more-cases-of-models-hiding-work-and-moving-data-they-should-not-have-movedOpenAI spent July explaining Hugging Face.Thu, 17 Sep 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/anthropic-found-three-cases-where-claude-hacked-real-companies-during-testshttps://replay.eclipticalabs.com/blog/anthropic-found-three-cases-where-claude-hacked-real-companies-during-testsA week after OpenAI admitted its evaluation agents reached Hugging Face, Anthropic published a review of its own cybersecurity evaluations. The review covered 141,006 runs.Fri, 31 Jul 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/openai-s-evaluation-agents-broke-out-and-hacked-hugging-facehttps://replay.eclipticalabs.com/blog/openai-s-evaluation-agents-broke-out-and-hacked-hugging-faceHugging Face disclosed an intrusion last week. OpenAI has now said the intruder was its own models.Wed, 22 Jul 2026 00:00:00 GMThttps://replay.eclipticalabs.com/blog/mixpanel-was-breached-openai-api-users-are-in-the-phishing-blast-radiushttps://replay.eclipticalabs.com/blog/mixpanel-was-breached-openai-api-users-are-in-the-phishing-blast-radiusOpenAI confirmed that a breach at analytics vendor Mixpanel exposed metadata about some users of its API. This is not a story about a model escaping a sandbox.Fri, 28 Nov 2025 00:00:00 GMThttps://replay.eclipticalabs.com/blog/anthropic-disrupted-an-espionage-campaign-that-used-claude-code-as-the-operatorhttps://replay.eclipticalabs.com/blog/anthropic-disrupted-an-espionage-campaign-that-used-claude-code-as-the-operatorAnthropic says it has cut off a campaign in which a Chinese state-sponsored group used Claude Code to run most of a cyber-espionage operation against about thirty targets.Fri, 14 Nov 2025 00:00:00 GMThttps://replay.eclipticalabs.com/blog/anthropic-says-a-hacker-used-claude-to-run-an-extortion-campaignhttps://replay.eclipticalabs.com/blog/anthropic-says-a-hacker-used-claude-to-run-an-extortion-campaignThat word does a lot of work. Read the mechanics before you file it under sci-fi.Thu, 28 Aug 2025 00:00:00 GMThttps://replay.eclipticalabs.com/blog/samsung-engineers-put-chip-making-secrets-into-chatgpt-no-one-hacked-themhttps://replay.eclipticalabs.com/blog/samsung-engineers-put-chip-making-secrets-into-chatgpt-no-one-hacked-themSamsung’s semiconductor division allowed ChatGPT in March. Within three weeks, internal reporting in Korea described three separate cases of employees pasting confidential material into the chatbot.Fri, 07 Apr 2023 00:00:00 GMThttps://replay.eclipticalabs.com/blog/a-bug-in-chatgpt-showed-people-other-users-chats-and-some-billing-detailshttps://replay.eclipticalabs.com/blog/a-bug-in-chatgpt-showed-people-other-users-chats-and-some-billing-detailsOn Monday, ChatGPT users started posting screenshots of conversation titles they did not recognize. By Wednesday, OpenAI had confirmed the obvious: this was not a rumor.Fri, 24 Mar 2023 00:00:00 GMT