65 researchers think that consent prompts cannot control AI agents
The authors want other groups to work on the 51 open problems in the report.
On 5 October 2026, Google Research released a report by 65 researchers. 36 of them are from Google. The authors think that consent prompts and static permissions cannot control AI agents that do tasks for users. They want each agent to check that each step is correct for the condition. They use the theory of contextual integrity. The report gives 51 open problems for other researchers.
Sources
Posted
