US government advisers have discussed how critical industries might report AI-powered cyberattacks, putting a growing security concern beside an approaching deadline for existing information-sharing protections. The discussion was reported on November 13, following a Thursday meeting of President Donald Trump's artificial-intelligence task force. It was a policy conversation, not an announcement of binding new incident-reporting rules.

According to Reuters, citing a Trump administration spokesperson, members considered how to encourage companies running essential services to disclose attacks and vulnerabilities without fearing legal repercussions. Defense Department technology lead Emil Michael, the task force's vice chair, also discussed work toward national standards for AI-developer safety-incident reports and cybersecurity information sharing.

Why the reporting question matters

An AI-enabled attack is not automatically a new category of cybercrime. The practical security question is whether automation changes how an event should be logged and shared. For example, defenders may need to preserve records of an AI agent's tool calls and permissions in addition to familiar network and account logs. That is an editorial recommendation, not a finding announced by the task force.

Reuters reported incidents involving agents moving outside testing boundaries. Those reports warrant scrutiny, but an incident described in testing does not establish that similar behavior is widespread in live critical infrastructure. The immediate policy challenge is more basic: an organization cannot help others defend against a new method if it keeps the relevant indicators, timing and affected systems private.

That matters for operators of power systems, banks and emergency services. Their security teams must decide what evidence can be shared promptly, which details are sensitive, and whether disclosures might expose customers or invite liability. Faster reporting may improve collective detection; compulsory reporting without clear safeguards might instead produce incomplete accounts or discourage early disclosure.

A December deadline, not a finished policy

The timing is important. The Cybersecurity Information Sharing Act of 2015 currently runs through December 11, 2026, under its extended statutory sunset provision. It provides legal protections for certain voluntary sharing of cyber-threat indicators and defensive measures. It is distinct from the Cybersecurity and Infrastructure Security Agency, which shares the CISA acronym.

Section 1503 of the US Code expressly authorizes certain monitoring, defensive measures and voluntary cyber-threat information sharing. This law is not itself a comprehensive AI-incident reporting framework. But any uncertainty around the legal basis for threat sharing can complicate efforts to persuade operators to disclose AI-assisted attacks. Extending existing protections and creating new reporting expectations would be separate policy choices, each requiring its own legal and operational details.

What remains unknown

In the November 13 Reuters account, the task force was discussing potential reporting arrangements, rather than announcing a final rule or implementation timetable. That report did not set out a specific list of covered AI incidents, organizations or disclosure fields. A useful standard would need to distinguish a hypothetical model vulnerability from an observed compromise, while explaining how sensitive details and personal data should be protected.

For businesses, a reasonable precaution is to maintain reliable incident logs, preserve evidence of AI-agent actions and define escalation contacts without assuming a new law is already in force. For policymakers, the test will be whether reporting rules produce actionable warnings without discouraging the very organizations expected to share them.

Sources

Reuters, November 13, 2026; US Code, 6 USC §1510 (effective period); US Code, 6 USC §1503 (sharing and monitoring).