AI news ·
Some federal agencies miss OMB deadline to manage risks from high-impact AI use cases
Several federal agencies missed the April 3 OMB deadline to implement risk safeguards for high-impact AI systems. Compliant agencies include Labor, NASA, and the VA; others reclassified AI uses or are still finishing steps.
Federal agencies miss OMB deadline for AI risk management practices
Most federal agencies met an April 3 deadline to implement risk management practices for high-impact AI systems, but several departments have not yet complied with the Office of Management and Budget requirement.
The Labor Department, NASA, the VA, State Department, GSA, and EPA completed their requirements on time. Other agencies reclassified their AI use cases or are still finishing implementation steps. A handful appear to have missed the deadline entirely.
The OMB memorandum requires agencies to either implement specific safeguards for high-impact AI systems or shut them down. High-impact uses include those that could affect civil rights, equal opportunities, or safety.
What agencies must do
Compliant systems must include:
- Pre-deployment testing and impact assessments
- Monitoring for adverse outcomes
- Human training and oversight
- Fail-safes that limit harm
- Appeal processes for affected individuals
- Mechanisms for user feedback
For managers implementing these practices, understanding AI for Government requirements helps clarify the specific compliance obligations. AI for Management resources also cover how to structure risk governance across organizations.
DOJ seeks major cybersecurity funding increase
The Department of Justice requested $149 million for its Justice Information Sharing Technology fund in fiscal 2027, a significant jump from the $38.5 million Congress appropriated in recent years.
The entire increase - $110.3 million - would fund DOJ's migration to zero-trust cybersecurity architecture for unclassified and national security systems. This approach treats every user and device as a potential threat and requires continuous verification.
DOJ cited funding cuts of $108 million in fiscal 2024 that have constrained its ability to defend its 275,000 endpoints and support 160,000 users. The agency said current funding levels are inadequate to address evolving threats.
The zero-trust model became standard practice across federal agencies following the SolarWinds attack in 2020, which compromised multiple government systems.