Larry Bradley is CEO & cofounder of SolasAI, an AI SaaS platform fixing model bias for Fortune 50 firms across finance, tech & healthcare.
In boardrooms everywhere, the conversations about AI are changing. What were once theoretical ideas and open experimentation have transformed into the real, the here and the now. In a short amount of time, AI (and by extension, AI enforcement) has become practice. And with that practice come new, more advanced discussions about the safety, efficacy and depth of this technology in our workplaces.
The urgency in this rapid shift of perspective is easy to understand. Just look at the adoption trends: In insurance, AI use jumped from 8% to 34% in a single year, up four times higher than a year ago. And yet many organizations still haven’t put clear rules in place for how their systems are used or governed. Deloitte found only one in five companies has a mature framework for overseeing autonomous or semi-autonomous AI. That means 80% of businesses are falling behind.
This gap between usage and governance matters even more when AI is increasingly being used in high-stakes workflows such as health insurance claims, lending and employment decisions. And when adoption outpaces governance, the cracks begin to show.
The Front Line Of AI Enforcement
The health insurance industry offers a view of where AI enforcement is headed. This year alone, Indiana, Utah and Washington all passed laws barring health insurers from using AI as the sole basis for denying claims. It’s creating a clear and growing consensus here in the U.S.: Fully automated denials cross a line, both legally and ethically.
And that consensus is having a real impact on our judiciary. A federal judge recently ordered UnitedHealth to hand over its internal AI-related files, signaling that companies may now be expected to produce records showing exactly how their automated systems were used and where humans remained involved in the process. After the ruling, some legal analysts now argue AI records should become standard discovery material in insurance litigation.
This could change the equation when it comes to your company’s policy on AI. “Human in the loop” can’t just be a slogan or a best-practice talking point anymore. It‘s now a core defense against legal intervention, but it only works if you have proof to back it up. If you can’t produce documentation on how a human reviewed, validated or overrode an AI-driven recommendation, you may quickly find yourself in trouble.
This doesn’t mean AI has no place in claims workflows or that you should abandon your AI plans. There are real efficiencies to take advantage of here. But the path forward is clear: We must retain the benefits of automation while ensuring AI is not the sole decision-maker in high-impact outcomes.
AI Enforcement Is Coming From The States
Much of the next phase of AI governance is going to be shaped at the state level, and that goes for both red and blue states. Texas as well as California now require human oversight for certain high-stakes AI decisions, a strong signal that enforcement of AI is not a partisan issue, and can’t be so easily avoided.
It’s also not just one or two isolated cases. More than 600 AI bills were introduced in state legislatures in the first quarter of 2026 alone, and over half of U.S. states have already adopted the insurance industry’s AI rule book in some form.
But laws are only part of the story. Enforcement matters just as much. Many consumers may not realize that state attorneys general are often the ones suing companies over biased or unfair AI systems. Roughly one-third of state AGs are on the ballot in the upcoming November midterms, and the outcomes of this election cycle will greatly influence how aggressively existing laws are enforced in the coming years.
There are already signs of what that enforcement may look like. In Texas, penalties tied to AI-related violations can reach $200,000 each. In New York City, restrictions on AI use in hiring and employment have pushed some companies to rethink their workflows nationwide.
There are even developments at the federal level. President Trump’s executive order requesting AI companies submit models for federal testing before deployment may give a clue about the White House’s intentions toward AI governance. The message across jurisdictions is consistent: AI doesn’t operate in a regulatory vacuum anymore.
AI Governance As Business Imperative
AI governance it table stakes now. AI is dominating companies’ IT spend, and governance is becoming a bigger portion of that spend as both investors and operators see it as a tool for faster, more effective adoption rather than a brake on innovation.
But governance is not a one-size-fits-all solution or a checklist to be marked off. Truly responsible systems must be flexible, dynamic and adaptable. Gartner predicts 40% of enterprises will abandon their AI agents due to governance failures, and in most cases it will be because these enterprises treated them like blanket solutions. Instead, we must implement proportional governance tools with different amounts of autonomy and trust, as well as “guardian agents” that support, monitor and protect our systems, capturing more AI value with less AI risk.
2026 is the year AI enforcement became real. The companies best positioned for this next phase of adoption will be the ones best integrating human-in and human-on the loop systems in tandem with agentic guardians—the ones who can explain, demonstrate and document their processes effectively. And that amount of oversight will end up leading to better outcomes for everyone.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
