The Pulse
Maryland Plans Frontier-AI Audits and 72-Hour Incident Reports
Maryland Gov. Wes Moore is directing the state AI Subcabinet to develop legislative recommendations for frontier AI companies. The proposal includes safety testing, independent audits, whistleblower protections, public reporting and 72-hour
AI.info Team ·
“With the speed with which AI is moving, Marylanders deserve guardrails that safeguard their rights, privacy, and dignity.”
Wes Moore, governor of Maryland
Maryland Targets Frontier Model Developers
Maryland Gov. Wes Moore’s administration is preparing legislative recommendations that would impose new oversight expectations on companies developing frontier artificial-intelligence models, including independent audits, public reporting and mandatory disclosure of serious incidents within 72 hours.
The proposal appears in an AI framework announced September 22. The governor’s office says the state AI Subcabinet will consult with outside stakeholders as it develops recommendations covering “frontier AI model companies.” The announcement does not present a finished bill, identify a threshold for which models would qualify, or set a timetable for legislation.
Moore described the package as a response to what his administration calls a lack of federal safeguards. “This technology is already here,” he said in the announcement, adding that Maryland could benefit from AI while acting to protect residents’ rights, privacy and dignity.
Audits, Testing and Public Reporting
The proposed framework groups several obligations under the regulation of frontier AI companies. They include safety frameworks and testing, independent third-party audits and evaluations, protections for workers, whistleblower safeguards, public reporting and 72-hour incident reporting.
The announcement does not specify what an audit would examine, who would qualify as an independent auditor, or whether companies would have to disclose complete evaluation results. It also does not define which incidents would trigger the 72-hour clock. Those details would have to emerge through the legislative process or subsequent regulation.
Maryland’s proposal would build on an existing state structure. The Artificial Intelligence Governance Act of 2024 established the governor’s AI Subcabinet and directed the Maryland Department of Information Technology to adopt safeguards for responsible use of AI in state government.
One Framework, Three Priorities
Moore’s announcement organizes the administration’s agenda around protecting Maryland residents, centering workers and keeping children safe. The frontier-AI provisions sit within the first category, alongside proposed action on algorithmic housing discrimination, AI-driven rent coordination, gambling operators targeting people showing problem-gambling patterns and the use of AI in workplace decisions.
The administration also proposes creating a statutory right of publicity so people would own rights in their likeness from the outset, rather than needing to prove harm after an image or identity has been cloned. Existing Maryland measures cited by the governor’s office address election deepfakes, AI-generated intimate images, AI impersonation and the use of automated tools in health-insurance utilization review.
Workers and Children Are Part of the Plan
The labor proposals call for a standing Labor-Management Committee involving state agencies, the AI Subcabinet and unions representing state workers. The administration also says it will work with lawmakers and stakeholders on protections related to social scoring, monitoring intended to weaken worker organizing and systemic bias.
For children, the framework calls for stronger protections around unsafe chatbots, limits on addictive design aimed at users under 16 and school protocols for AI tools that students interact with in classrooms. Moore’s administration also proposes a “do no harm” policy requiring classroom AI tools to show educational benefits, supplement teacher instruction and be disclosed to families before use.
What the Announcement Leaves Open
Moore’s plan is a policy direction, not yet an enforceable frontier-model law. The governor’s office has not released draft statutory language, a covered-company list, penalties, audit standards or a process for regulators to review confidential model information.
That leaves the next phase with the AI Subcabinet, lawmakers and the stakeholders named in the announcement. Maryland already has experience regulating specific uses of automated systems, including insurance decisions and state-government deployments. The frontier-AI proposal would extend that work toward the companies that build the most capable models, but its practical reach will depend on the legislation the administration ultimately sends to lawmakers.