AI policy across the United States is moving fast and in multiple directions. In the same month that the White House sent Congress its AI legislative framework, states like California and Pennsylvania are advancing their own legislation. For companies building AI tools for government, this poses an operational nightmare, and for the government agencies wanting to modernize, this slows down procurement for the AI tools needed to achieve the much-needed digital transformation.
The Road to AI Adoption In Governments
The mission for local, state, tribal and federal Governments is to provide meaningful services and outcomes for its constituency. As a virtue of this mission, governments need to properly allocate resources in order for its staff to be able to execute.
Corporations have a fiduciary responsibility to properly allocate resources in order to achieve its mission of maximizing shareholder value and generating a return on investment. Corporations, because of their mission, adopt specific processes and technologies in order to optimize their ability to deliver its mission.
Governments have a completely different set of obligations than corporations: their obligation to their constituency is that the technologies they adopt are secure, enduring systems. In order for governments to purchase technology, beyond security, a budget justification and an official procurement process is required. The process is cyclical and often requires approval from multiple positions across teams and divisions within a government. Because of this, there’s often a need to find vendors that have the largest capacity and tools to solve the largest amount of problems rather than procure a dozen independent solutions.
Risk vs. Reward
Frontier large language models and the applications that are powered by them, generally referred to as artificial intelligence, have the potential to accelerate governments’ mission to a level comparable to corporate efficiency. This idealistic outcome can only be achieved through aggressive innovation, deployment and adoption of government-serving AI technologies.
Those most likely to develop solutions that governments want to adopt in order to modernize could face material friction through patchwork state AI laws. It is absolutely necessary and critical that States protect their residents, children and communities from malicious, evil AI actors. The other side of the argument is that preemptive legislation has the potential to over-regulate innovators.
When Compliance Complexity Outpaces AI Innovation
The risk of patchwork state legislation is that it can hold those with the least resources (innovators of AI technology, typically startups and small businesses)to a degree of compliance that’s not sustainable. Larger corporations and labs, armed with resources and expertise would be the select few that could navigate the compliance and reporting requirements needed to abide across the 50 states.
Governments that want to adopt AI solutions, typically developed by startups and small businesses, would have to ensure that their procurement process follows state laws. The downstream effects of this model are that startups would have to set up compliance review and reporting processes, limiting scale, before being able to move on to the next.
Is A Federal Framework The First Step?
A first-step to reduce complexity allowing governments to more seamlessly adopt AI, are federal uniform guidelines, similar to the NIST guideline: a comprehensive set of standards and best practices designed to secure information systems against cyber threats. The framework assists organizations with risk management, security control implementation, and compliance, covering topics from data protection to incident response.
A similar framework for artificial intelligence could provide local, state and tribal governments a uniform picture of best practices they should hold AI vendors to. This framework could be further narrowed down by industry. In government-technology, it could offer startups a way to build with expected guardrails, while providing governments, wanting to modernize using AI, a clear basis for evaluating potential vendors through its procurement process.
While this would only be a first step, it enables a path for government to modernize while giving AI startups best practices to adopt early. The framework can mirror expected legislation at the federal level and serve as an experiment in what proper uniform federal legislation would create.
The Opportunity Ahead
Governments are entering a period where they’ll be able to modernize as fast as their contemporary organizations. The only way for innovators, dreamers, and builders to have the proper environment to build and deploy systems, is with a proper guidance set in place. A federal framework offers exactly that, while giving government agencies seeking to adopt the technology a clear picture of best practice. Without this in place, state legislation, especially conflicting patchwork legislation could skew to primarily benefit only the largest AI organizations. These organizations would be the only players who have the resources and teams capable of handling such requirements.
The smaller players, those able to solve the real rapid needs of the government, would be unable to compete. AI startups are often best positioned to address immediate needs within government. Specifically solutions addressing needs that require rapid development and release. Startups generally have fewer internal layers of administration meaning they can move from ideation to prototype to deployment quickly.
In a fragmented regulatory environment, however, AI startups would need to allocate material resources toward complying state level requirements. This introduces complexity, startups can’t manage and would disproportionately affect smaller teams, reducing time and capacity for product development.
Larger corporations can absorb changing requirements much faster with larger teams and greater levels of resources to allocate. Their scale can support dedicated compliance teams. But it’s also because of their large resource allocation and team sizes, corporations are not able to ideate, prototype and deploy like startups. Federal guidance migrating into federal legislation creates a better ecosystem to sustain both rapid development and deployment of iterative solutions and larger more technical and mature initiatives to grow.
