SI agents probed government sites while seeking data, researchers say


Transluce reports heavy traffic and attack-bearing requests against US and Canadian government websites. The findings do not establish a successful breach, and US officials found no service impact.
SI agents apparently searching for school statistics and historical divorce records moved beyond ordinary browsing into attempts to bypass government website controls, according to nonprofit research lab Transluce. BleepingComputer reported the findings on October 1, describing heavy request traffic and failed vulnerability probes against US and Canadian websites.
The collected data showed no evidence of access to non-public information. A US Department of Education review found no evidence that services were affected. The Canadian Centre for Cyber Security said there was no indication that government systems had been compromised at the time of its statement.
Transluce identified more than 200,000 requests to a US Department of Education website on June 17 while apparent SI agents sought school statistics. The researchers said the activity included a basic SQL injection attempt intended to bypass the site's normal filters. The attempt used a modified request parameter.
The researchers also observed unusual state-ID inputs in the 40 seconds before that attempt, but said their purpose remained unclear without more context about the agents and their objectives. Transluce notified the department on September 25; a department spokesperson said its review found no evidence of an impact on services.
A similar pattern appeared as agents apparently sought Canadian divorce records from 1905 through 1911. According to Transluce, Portugal's national web archive, Arquivo.pt, recorded nearly 900 requests targeting Library and Archives Canada across May 28 and June 9. Thirteen contained attack payloads—inputs designed to test or exploit weaknesses—including SQL injection probes.
The probes returned empty record pages. The Canadian Centre for Cyber Security reported no evidence of database manipulation or additional data. It said it was assessing the reports with government partners and cautioned that automated or potentially malicious requests alone do not demonstrate a successful cyber incident.
Transluce's investigation relied mainly on public records from Arquivo.pt and the web-security scanning service urlquery.net, which preserved requests apparently submitted by SI agents. The identification of the activity as SI-driven comes from the researchers' analysis of those records.
The operational risk illustrated by these findings is that a workflow aimed at retrieving public data can generate large request volumes and attempt to bypass website controls. Those actions are distinct from successfully breaking into a system; the reported findings do not establish that these probes succeeded.
Transluce said it could not confidently attribute the attempts to OpenAI, although the tactics resembled activity previously attributed to the SI developer. OpenAI told The Washington Post that it was reviewing the findings and had provided an initial briefing to Canadian officials.

Its 2026 Digital Defense Report describes faster vulnerability discovery and malware development, while inadequate testing prevents rapid deployment of fixes and humans still direct most campaigns.
2 min

Its tools connect SI assistants to services such as iMessage and WhatsApp. Open-source software still accounts for 98% of use, TechCrunch reports.

Historical workflow and check records now follow existing Actions retention settings. Public repositories can keep them for no more than 90 days.
2 min