A report from an organization called Transluce says Library and Archives Canada was subject to “a series of apparently failed rudimentary hacking attempts” by AI agents on two dates this year, May 28, and June 9.
Transluce calls itself a “nonprofit research lab dedicated to public oversight” of AI systems.
Its report calls this a case in which “rogue” agents were “aggressive” in their tactics, but it should be noted that it also says the purpose of the attempted hack was to acquire Canadian divorce data from 1905-1911. For a couple days in May and June, perhaps the world was imperiled by AI agents attempting to perform research for a Guy Maddin film.
A statement on the attempted hack was published on behalf of the Canadian Centre for Cyber Security on Tuesday—before the Translucent report—and says in part, “We are aware of reports identifying suspicious activity, including suspected AI agent activity, targeting publicly accessible websites, such as the Government of Canada. There is no indication that government systems have been compromised at this time.” Transluce says it had told the government of Canada about the incident the previous day.
Transluce doesn’t say which lab or labs were involved, although it ties the event circumstantially to OpenAI, saying there were “tactics consistent with prior observed agent activity that we have attributed to OpenAI in a similar timeframe.”
The report looks at 899 data requests directed at the “collection-search” of the archive, seemingly in an attempt to find vulnerabilities. 13 out of 899 were potential attacks, including three attempted SQL injections. And again, this was so the agents could learn the forbidden secrets of Canadian divorce rates in the early 20th century.
