Anthropic said the review did not find any “incidents of similar severity to the cybersecurity incidents we reported this summer.” The review then turned to transcripts from “lower risk” test scenarios, such as the one affecting the Philadelphia Police Department.
“While the impact of these behaviors was minimal and we had already disabled live internet access for some high-risk and cybersecurity assessments, we have now decided to expand that to include all of our internal assessments until we have confirmed that our security and monitoring measures… reliably detect behaviors like these,” Anthropic said.
Anthropic says new tools block unwanted behavior
Anthropic said it stopped running some assessments and moved others to offline versions “or rebuilt them so that their tasks do not reach live websites.” The company “updated the security barriers on some of our Internet access tools, such as the web search tool, to greatly restrict what the model can do with them,” he said.
Anthropic said it also created new tools “to automatically detect and block the types of behaviors described above… When we tested it with the cases described in this post, it blocked them all.”
The White House AI task force, which it calls the Super Intelligence Force, issued a statement saying it informed Anthropic “that we expect immediate and full transparency to the entities involved and the public.”
The White House AI Task Force also said that Anthropic must “immediately provide remedial services to affected entities and any harmed Americans,” and that all AI companies “must immediately disclose incidents involving their models and follow up with swift and decisive action to remedy any and all harm,” according to an Axios report.
