AI Risks in Government
Due-process challenges, OMB M-24-10 compliance, FOIA cases, and accountability actions — scored from public records.
Industry overview
AI in government carries due-process, equal-protection, and transparency obligations that private deployments do not. OMB M-24-10 sets minimum practices for federal use of AI in rights- or safety-impacting contexts. State and local agencies face increasing algorithmic-accountability legislation. FOIA and state public-records laws apply to AI vendor contracts and outputs, and several agencies have lost cases attempting to redact them. The cost of an unconstitutional or non-transparent deployment is borne by the public, then by the agency.
Key risks for Government
Due-process and equal-protection challenges
Algorithmic decisions affecting benefits, parole, child-welfare, or housing have been challenged on constitutional grounds. The standards are evolving, but the trend is toward requiring meaningful explanation, contestability, and bias review.
OMB M-24-10 and federal compliance
Federal agencies must inventory AI use, designate Chief AI Officers, and implement minimum practices for rights- and safety-impacting AI. Non-compliance is increasingly visible to oversight bodies and inspectors general.
Public-records and FOIA exposure
AI vendor contracts, model documentation, and prompt logs are typically public records. Vendor confidentiality clauses do not, by themselves, defeat disclosure.
Procurement and security baselines
Federal procurement (FAR/DFARS) and FedRAMP requirements apply to AI services. Many AI vendors do not meet the baseline expected of cloud providers in regulated agency environments.
Regulatory surface
Regimes: U.S. Constitution (due process, equal protection), OMB M-24-10, state algorithmic-accountability statutes, FOIA and state public-records laws, FedRAMP, FAR/DFARS, EU AI Act for cross-border use.
AI services tagged for Government
23 servicesGrok
34LLM / Chatbots
ModerateIncidents in Government
- AI Incident Database: Criminal investigation launched after Swiss minister’s chatbot misogyny complaint2026
- AI Incident Database: Chatbot d’IA:enquête pénale après la plainte de Karin Keller-Sutter2026
- AI Incident Database: Wegen Beleidigungen: Keller-Sutter geht juristisch gegen Elon Musks KI Grok vor2026
ChatGPT
34LLM / Chatbots
ModerateIncidents in Government
My AI by Snap (rebranded)
32AI Agents
ModerateIncidents in Government
Snap My AI
32AI Agents
ModerateIncidents in Government
Clearview AI
25Enterprise AI
ModerateIncidents in Government
NSO Group Pegasus
10Other
LowIncidents in Government
Scale AI
7Enterprise AI
LowIncidents in Government
IDEMIA MorphoFace
1Other
LowIncidents in Government
Buyer checklist
- 1
Documented assessment of whether the deployment is rights- or safety-impacting under M-24-10.
- 2
Public-records and FOIA-readiness review of the vendor contract and operational artifacts.
- 3
Bias and disparate-impact testing that satisfies the agency's constitutional and civil-rights obligations.
- 4
Contestability path: how does an affected person challenge an adverse algorithmic decision?
- 5
Procurement and security baseline that meets, not approximates, the regulated-agency standard.
Frequently asked
Does OMB M-24-10 apply to my agency?▾
It applies to federal executive-branch agencies subject to the Federal Information Security Modernization Act, with carve-outs for the IC and certain national-security functions. Many state and local agencies are adopting analogous frameworks.
Are AI prompt logs subject to FOIA?▾
Generally yes, when the prompts and outputs are records used in agency business. Vendor confidentiality assertions do not override FOIA — exemptions must apply on the record's own terms.
Get alerts when Government risk scores change.
Court cases, breaches, and regulatory actions — pushed to you when they affect this industry.