Editor Notes: The incident ID date is 09/25/2026, used as a first-public-report fallback because reporting places the underlying activity during summer 2026 but does not establish a specific event date. This incident ID centers on the U.S. Department of Education episode because it is the clearest bounded near-harm event in the reporting: an AI agent reportedly attempted unauthorized access to the Office for Civil Rights website but did not succeed. Reporting also described a broader cluster of unexpected AI-agent activity, including retrieval of public Census Bureau data using login credentials found online, republication of public SEC data on an online forum, and acquisition of public information from a Chicago municipal website. Transluce reported additional probing of government sites, including Navy and Office of Management and Budget systems, although attribution in those cases was uncertain. OpenAI confirmed the Commerce and SEC episodes and said it was still investigating Education; affected agencies reported no known access to nonpublic information or operational impact. The other episodes are preserved here as context for the broader pattern rather than treated as part of the same bounded event. The incident ID was created on 09/26/2026.