State and local governments are rapidly purchasing AI systems, but their contracts overwhelmingly lack basic protections for transparency, fairness, and accountability. A comprehensive analysis of over 1,000 AI contracts across California, Utah, and Florida found that while 77% of contract provisions are standard boilerplate language, only 5.3% address transparency, 3% address cybersecurity, and just 2.4% address fairness and accountability. This governance gap is particularly concerning because these procurement decisions lock in policy choices for years, with some contracts spanning a decade or longer.

Why Should Governments Care About AI Contract Language?

Procurement is not simply an administrative task; it is the primary mechanism through which AI enters government and the first line of defense for responsible AI deployment in the public sector. When contracts lack robust safeguards, they can hard-code opacity, vendor lock-in, and weak accountability for years or decades. The median state AI contract is valued at approximately $1 million, meaning poor contract terms affect significant public resources and, more importantly, the citizens who depend on these systems.

The consequences of inadequate contract language are not theoretical. Michigan's experience with its unemployment insurance system illustrates the real-world harm. From 2013 to 2015, the state's AI-powered fraud detection system, called MiDAS, wrongly accused more than 34,000 unemployed individuals of fraud. The original $47 million contract with Fast Enterprises and the subsequent $78 million replacement contract with Deloitte neither included meaningful provisions for algorithmic transparency, bias testing, or independent performance auditing. The MiDAS debacle ultimately cost the state over $125 million across two contracts, falsely accused 40,000 residents, and resulted in a $20 million class-action settlement.

What Are the Key Gaps in Current State AI Procurement?

Government agencies typically evaluate AI purchases based on cost, vendor qualifications, and compliance with existing regulations, but they rarely assess algorithmic risk or ask critical questions about how the systems actually work. Agencies may not inquire about bias testing, government access to training data, or vendor requirements to disclose how the model makes decisions. This knowledge gap exists because procurement capacity has not kept pace with technical complexity, leaving many agencies ill-equipped to evaluate performance, negotiate price and scope, and ensure these tools are used effectively and responsibly.

The problem is compounded by the widespread use of cooperative purchasing agreements, in which one state competitively bids and negotiates a contract that other states and local governments can adopt without rerunning the procurement process. More than 4 out of 5 state AI contracts in the EPIC (Electronic Privacy Information Center) dataset were negotiated through the NASPO ValuePoint platform, a cooperative contract program. While this approach saves time and resources, it also concentrates risk; boilerplate language from the initial contract often becomes the template for all participating jurisdictions.

How Can States Strengthen AI Procurement Practices?

Experts recommend three concrete reforms that states can implement without waiting for new legislation:

• Standardized Responsible AI Contract Clauses: States should adopt contract language aligned with the NIST AI Risk Management Framework, ensuring that all AI procurements include provisions for transparency, fairness, and accountability from the outset.
• Risk-Tiered Procurement Review Processes: Governments should implement tiered review processes modeled on proven approaches in San José and Colorado, where the level of scrutiny matches the potential impact of the AI system on citizens.
• Mandatory AI Vendor Fact Sheets: Vendors should be required to provide standardized fact sheets disclosing key information about their AI systems as a condition of contract award and renewal, enabling agencies to make informed decisions.

These reforms address a critical insight: contract language is often a relatively low-friction and politically viable tool that can generate concrete governance benefits without requiring new AI legislation. Procurement decisions made today will shape how AI is used in criminal justice, healthcare, education, and other consequential domains for years to come.

What Does This Mean for the Future of Government AI?

The race between technological change and government capacity is accelerating. EPIC documented 600 state AI contracts in 2023, and analysis of just three states identified over 1,000 contracts. As governments continue to adopt AI at an increasing pace, the decisions made in procurement offices will determine whether these systems operate with appropriate oversight or whether they repeat the failures of systems like MiDAS. The stakes are high: inadequate contract provisions do not merely waste public resources; they can cause direct harm to the people these systems are meant to serve.