Above Security calls for independent evidence and human oversight of AI-assisted security investigations. In their own October 10, 2026 posts, CEO Aviv Nahum and Chief Product and Technology Officer Amir Boldo back Satya Nadella's proposal to control AI models like privileged insiders. They explicitly apply the requirement to AI systems investigating other agents.
Nahum argues that technically permitted access does not establish that an action fits the assigned task. An agent renewing a customer contract may need different data from a coding agent. Investigators would need to connect the task, identity, affected information and sequence of actions to assess deviations. His examples describe possible situations, not reported incidents.
Boldo separates three responsibilities: what independent systems recorded, how an investigative model interprets that information, and who may act on the findings. A model-generated explanation should remain checkable against a reviewable timeline and the underlying observations. An AI investigator's interpretation is not automatically an established fact about someone's behavior or intentions.
The CPTO describes revoking credentials, restricting access and pausing an agent as decisions that independently operated safeguards would enforce. Nahum calls for Nadella's principles to become enterprise security requirements. The authors also reference a framework developed with Forscie for investigating AI agents; the posts do not establish a new launch date for it.
Both texts connect their security position with their own product business. Architecture and integration descriptions are vendor claims, not independent effectiveness measurements. The posts report neither newly verified protection results nor a specific deployment date. They establish the authors' own substantive support and their call not to make investigative AI the sole authority over evidence and enforcement.