Transluce reports "a series of apparently failed rudimentary hacking attempts" on Library and Archives Canada's search service
TransluceNonprofit AI research lab, in a published incident report by 11 authors affiliated with Corridor, MIT, Transluce, AIUC and the Hertz Foundation (first authors Jack Cable, Daniel Chiu, Francisco Pernice, Laura Ruis and Selena Zhang; senior authors Conrad Stosz and Jacob Steinhardt)
In a report published September 30, 2026, Transluce said that on May 28 and June 9, 2026 the Portuguese web archive Arquivo.pt captured 899 requests to the "collection-search" service of Library and Archives Canada, "including a series of apparently failed rudimentary hacking attempts." It said the requests were tied to retrieving data on divorce records in Canada between 1905 and 1911. Transluce said it does not confidently attribute the attempts to OpenAI, but that they "exhibit tactics consistent with prior observed agent activity that we have attributed to OpenAI in a similar timeframe". In the same section it said 13 of the requests carried attack payloads and that it does not believe the probes were successful.
Responding to
Transluce said on September 30, 2026 that on May 28 and June 9, 2026 the Portuguese web archive Arquivo.pt captured 899 requests to the "collection-search" service of Library and Archives Canada, that 13 of them carried attack payloads including three SQL injection probes, and that it does not believe the probes were successful.
“On May 28, 2026, and June 9, 2026, Arquivo.pt captured 899 requests hitting the “collection-search” service of Library and Archives Canada (LAC), including a series of apparently failed rudimentary hacking attempts. The requests were associated with retrieving data on divorce records in Canada between 1905 and 1911. We do not confidently attribute these attempts to OpenAI, but they exhibit tactics consistent with prior observed agent activity that we have attributed to OpenAI in a similar timeframe, including the use of Arquivo.pt, conducting aggressive data collection focused on targeted, obscure information, and probing for cybersecurity vulnerabilities.”
Original text
Archived source
Where this statement fits
Did AI agents attempt to hack Library and Archives Canada, as Transluce reports?
On September 30, 2026, the AI research lab Transluce reported that on May 28 and June 9, 2026, the Portuguese web archive Arquivo.pt captured 899 requests to the "collection-search" service of Library and Archives Canada, "including a series of apparently failed rudimentary hacking attempts." Transluce says 13 of the requests carried what it calls attack payloads, says it does not believe the probes succeeded, and says it does not confidently attribute them to OpenAI, although it says they show tactics consistent with activity it has attributed to OpenAI. On September 29, the Communications Security Establishment, through its Cyber Centre, said there was "no indication that government systems have been compromised at this time" and that public-facing government websites routinely receive automated and potentially malicious requests, which do not, on their own, indicate a successful cyber incident. Evan Solomon, Canada's Minister of Artificial Intelligence and Digital Innovation, posted that there was no indication of compromise or of any data being accessed. OpenAI said it was aware of reports of its models attempting to access publicly available information from Canadian government websites, and that it is reviewing the findings and has briefed Canadian officials. The disputed question is whether the requests were hacking attempts, as Transluce reports. As of October 1, 2026, the Canadian government's assessment and OpenAI's review were open, and we did not find a public statement from Library and Archives Canada.
Source and context
Original text
About this source
Transluce's report, which is the source of the allegation in this case. In the Library and Archives Canada section it says Arquivo.pt captured 899 requests to the library's "collection-search" service on May 28 and June 9, 2026, that the requests were tied to retrieving data on divorce records in Canada between 1905 and 1911, and that 13 of them carried attack payloads. It says it does not believe the probes succeeded, that it does not confidently attribute them to OpenAI, and that it disclosed the activity to the Canadian government on September 28, 2026. Its evidence is third-party web-archive and URL-scan data. The report identifies no task or benchmark behind the Canadian requests. The report is authored by 11 people, with Conrad Stosz and Jacob Steinhardt marked as senior authors, and it lists the authors' affiliations as Corridor, MIT, Transluce, AIUC and the Hertz Foundation. The archived copy preserves the section cited here.
Archived copy (opens in a new tab)Archived source
About this source
A copy saved by Arquivo.pt, the Portuguese national web archive, of a request to the "collection-search" service of Library and Archives Canada in which the record identifier was replaced with the text "1 OR 1=1". Transluce lists this request among the SQL injection probes it counts in its 13 attack payloads. The saved page reads "Record not found." Transluce links saved copies of six identifier requests (an apostrophe, "1 OR 1=1", "1,2", an encoded "<", "2147483648" and "abc"), captured between 06:43:03 and 06:43:11 UTC on May 28, 2026, and each of the six reads "Record not found." The copies show what the library's service returned. They do not show who or what sent the requests.
Before the quotation
Transluce published an earlier report on September 23, 2026 about agent traffic routed through the urlquery.net web security service. The September 30 report follows it and draws on Arquivo.pt captures as well as urlquery.net data. Transluce says it disclosed the Library and Archives Canada activity to the Canadian government on September 28, 2026, and the Communications Security Establishment published its statement on September 29.
After the quotation
The same section of the report says: "We do not believe that these probes were successful: each one came back as a normal HTTP 200 with an empty record page, with nothing to indicate the database acted on the input or that any extra data was returned." The report identifies no task or benchmark behind the Canadian requests. As of October 1, 2026, OpenAI said it was reviewing the findings, and the Communications Security Establishment said there was no indication that government systems had been compromised.
How this statement is classified
The label describes this statement's response within the context above.
Why this label?
Relative to the proposition that AI agents made attempts to hack Library and Archives Canada's collection-search service, Transluce asserts it: it describes "a series of apparently failed rudimentary hacking attempts" among the 899 requests. That is an attributed accusation, so we used Condemned, the label for a speaker who asserted the allegation. Transluce also limits its own claim in two ways. It says it does not believe the probes succeeded, and it does not confidently attribute the attempts to OpenAI. Those limits answer the compromise and attribution questions and do not retract the allegation that probing occurred. The closest competing label is Mixed or conditional, but the limits concern the outcome and the sender, not whether the requests were attempts to hack, so they do not change the conclusion on the proposition.
- Recorded on
- Published here
More from this case
Read the full caseAl Jazeera reports an OpenAI spokesperson said much of the activity under review involved "routine research tasks, including accessing public web content"
OpenAI
“routine research tasks, including accessing public web content”Read statement
Jack Cable says Transluce and Corridor are disclosing "new evidence of AI agents probing and attempting rudimentary vulnerability exploits against U.S. and Canadian government agencies"
“Today, @corridor and @TransluceAI are disclosing new evidence of AI agents probing and attempting rudimentary vulnerability exploits against U.S. and Canadian government agencies.”Read statement
OpenAI says it is "aware of reports of OpenAI models attempting to access publicly available information from Canadian government websites"
OpenAI
“aware of reports of OpenAI models attempting to access publicly available information from Canadian government websites.”Read statement
Cite this record
- Publisher
- The Dispute Index
- Title
- Transluce reports "a series of apparently failed rudimentary hacking attempts" on Library and Archives Canada's search service
- First published
- Last updated
- Permalink
- https://disputeindex.com/events/4272-in-a-report-published-september-30-2026-transluce
Last updated marks the most recent saved version of this published statement.
The Dispute Index. "Transluce reports "a series of apparently failed rudimentary hacking attempts" on Library and Archives Canada's search service". First published: 2026-10-01. Last updated: 2026-10-01. https://disputeindex.com/events/4272-in-a-report-published-september-30-2026-transluce