Diving Headfirst Into The Google Newly Released ‘AI Governance In America’ Framework

AI And The Law

As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the intersection of AI and the law for many years. You can find my writings not only in my Forbes column but also as posted in Bloomberg Law, ABA Law Journal, The National Jurist, The Global Legal Post, Lawyer Monthly, The Legal Technologist, MIT Computational Law Journal, and so on.

There are two major perspectives on the mixture of AI and law:

• (1) Law & AI. The application of laws to the governance and regulation of AI.
• (2) AI & Law. The application of AI to perform legal reasoning.

Thus, you can apply the law to AI, and conversely, you can apply AI to the law. For my big picture overview of both of these exciting and rapidly evolving realms, see my discussion at the link here and the link here.

When it comes to applying the law to AI, the aim is to establish suitable regulations and provide appropriate governance on how AI should be devised and implemented. There are longstanding concerns that AI makers aren’t giving due attention to the ethical ramifications of their wares. Ethical issues are construed as “soft laws” and aren’t as formidable as legally enacted laws, known as “hard laws”. To level the playing field and keep AI makers on the up-and-up, some believe that we need more AI laws.

On the other side of the coin is the application of AI to the law. This consists of using AI to aid legal activities. Lawyers tap into the latest AI to devise legal strategies, brainstorm to find creative legal arguments, draft court filings, and prepare for cases by having the AI pretend to be an able adversary. For my extensive coverage on AI for legal reasoning (AILR), see the link here.

The Current Situation Legally

In terms of the AI laws in the United States, they have not yet stood the test of time, meaning that we won’t really know how well they stand up until there are court cases that test these new laws. It is too early to know whether the laws will survive legal battles waged by AI makers and other contenders. Just because AI laws are enacted does not mean they are proper. All sorts of improper provisions and constitutionally contentious stipulations are undoubtedly buried within these shiny new AI laws.

Congress has repeatedly waded into establishing an overarching federal law that would encompass AI. So far, no dice. The efforts have ultimately faded from view. Thus, at this time, there isn’t an overarching federal law devoted to these controversial AI matters. The big question will be to what degree a sweeping federal law would impact the numerous state-level AI laws. The odds are that many of the state-level laws would run afoul of a federal mandate, and a tsunami of legal cases would arise as a tussle between federal law and state law is undertaken. It surely will be a legal mess.

The crux is that there is intense and pervasive interest in using the law to govern AI. It is an abundantly burgeoning realm. AI companies would be wise to keep a close eye on what is happening in the hallways and byways of regulators and legislative bodies. I have repeatedly noted that a profitable specialty for budding lawyers is to consider concentrating on the exciting and dynamic field of AI and the law; see my predictions and suggestions at the link here.

Frontier AI Is The Aim

The advent of generative AI and large language models (LLMs) has spawned a now-common phrase for being at the leading edge or frontier of AI. When an AI maker comes out with their latest LLM, it is said to be at the cutting edge or frontier of where AI is heading. These frontier models tend to incorporate brand-new foundational advances in AI; the models are usually massive in size and require vast amounts of computing to produce. A simple analogy is to liken this to cars, whereby one might say that an automaker has come up with a new, fully loaded car that has all the latest advances and capabilities.

The policy issue is whether the maker of a frontier AI model should be required to undergo some kind of legally mandated federal pre-checks and testing before the AI maker can release the AI to the public or even for private use. One argument in favor of such legislation is that an AI maker could otherwise freely unleash an AI that is going to wreak havoc. In the car analogy, you might say that it would be akin to having no federally mandated requirements to pre-test cars for safety and reliability.

AI makers are in a free-for-all and can currently do as they wish. Concerns about the unabated release of frontier AI models came to the fore when a recent frontier model was found to have discovered cybersecurity hacks that could potentially undermine computer systems everywhere; see my coverage at the link here. The AI maker opted to wait to release the frontier model. Ought that choice be up to the AI maker, or should Congress pass a federal law that requires frontier models to undergo strictly stipulated pre-checks before they can be released?

A contention made by some is that federally mandated pre-checking would dampen and slow the advancement of AI in the United States. Other countries that don’t force such pre-checks would proceed ahead of us. We would be shooting our own foot in the high-stakes race of AI progress. The counterargument is that we could potentially gut our own country by allowing our AI to go unchecked into the marketplace. Others argue that there can be a middle ground that serves as a Goldilocks approach, namely that the porridge is not too cold or too hot, and that federal laws could be crafted to achieve a balance of risk versus reward.

For more on my analysis of this topic, see the link here.

Google Offers Its Middle Ground

Speaking of a middle ground, Google recently released an AI policy document entitled “A Pragmatic Approach to AI Governance in America” (June 2026) that they assert provides a sensible and practical middle-ground approach to “regulating” frontier AI. To clarify, this doesn’t identify specific new AI laws that might be enacted but instead focuses on the creation of a national entity that would have responsibility for overseeing frontier AI in the U.S.

Here are some excerpted salient points from the AI policy proposal:

• “Just as there isn’t a single question about AI, or a single goal policy can achieve, there is no single answer to what AI policy should be. In this paper, we build on the best ideas we have seen to separately address the national security risks of frontier AI and the economic and social impacts of widely deployed AI.”
• “To protect American innovation and ensure global AI leadership while providing for a safe and secure digital future, leading labs need a unified framework for frontier AI safety, security, incident reporting, and transparency.”
• “We suggest federal policymakers consider a frontier AI regulatory organization (FARO). A FARO could progress and promote national and international standards, guiding requirements for how developers should identify and mitigate risks and verifying that companies implement security practices and incident response plans before releasing frontier models publicly.”
• “The issues raised by the widespread use of AI applications like chatbots are distinct from the kinds of national security issues posed by advanced frontier AI models. The U.S. federal government should also, but separately, address everyday uses of AI across the economy through a series of discrete frameworks.”

I will walk you through some of the aspects that have particularly drawn controversy or have at least gotten a lot of chatter online. One thing to realize is that this isn’t some far-fetched or out-of-the-blue proposal; namely, there have been others calling for establishing an entity of this nature. The mainstay is that this is Google making this recommendation and therefore carries perhaps hefty weight in comparison to AI researchers or others who have been doing likewise.

Debating The Degree Of AI Regulations

One notable grievance is that there seems to be a new vibe going around overall that attempts to couch the AI regulatory gambit somewhat perniciously. This Google proposal appears to play into that vibe. Allow me to lay this out for you.

We have so far been mired in the on/off dichotomy debate, consisting of two starkly contrasting possibilities:

• (1) No regulations on AI.
• (2) Heavy regulations on AI.

You can imagine that this is the typical polarization we seem to have in our society today. Things are either one way or the other way. There is no room in between. To try and get above this earthly dichotomy, many are now saying they are tired of the on/off earsplitting disputes and wish to offer a third option. This certainly sounds refreshing.

The third option is nearly always placed smack dab in the middle of the other two:

• (1) No regulations on AI.
• (2) Middle-ground regulations on AI.
• (3) Heavy regulations on AI.

The beauty of this clever positioning is that anyone taking this stance appears to be above the fray. You would thus fall into the mental trap that they must be taking a balanced approach. It is a trick of taking your mind away from the distasteful extremes of debate and appearing to offer a reasonable compromise, which maybe it is or maybe it isn’t.

To show you what I mean, we could split the apple pie into four slices rather than three, such as this:

• (1) No regulations on AI.
• (2) Modest regulations on AI.
• (3) Substantial regulations on AI.
• (4) Heavy regulations on AI.

The gist is that we now have a middle ground that consists of two possibilities. The other middle ground in the three-count setup could have been closer to no regulations or closer to heavy regulations. We don’t really know that it is somehow purely in the middle. It appears to be in the middle simply because it is the third option that happens to be between the other two extremes.

Of course, we can keep going in this splitting of the pie:

• (1) No regulations on AI.
• (2) Minimal regulations on AI.
• (3) Modest regulations on AI.
• (4) Substantial regulations on AI.
• (5) Heavy regulations on AI.

Claiming that something is a middle ground is not necessarily a middle ground in the sense of being perfectly positioned as the middle stance of two extremes. We must be careful in falling for semantic wording that instantly gets us to perceive a so-called middle ground as the better or best option. I say this due to the recent tendency for lots of pundits at the extremes who are now claiming their opinion is the middle ground, doing so by either making this up or by replacing the goalposts with extremes that seem to cast their posture into a middle ground.

Focus Of AI Regulations

Another crucial consideration is that there is a trend to divide up AI into two types, consisting of frontier AI and non-frontier AI. Some refer to the non-frontier AI as being below frontier AI. In that sense, we are to think of frontier AI as the tiptop advanced AI, and then anything else is accordingly below that top level. If we merely said non-frontier AI, presumably this could suggest that there is other AI above the frontier AI.

Not everyone is comfortable with dividing AI into those two types. The Google proposal opts to use that framing, such that the FARO would only be overseeing frontier AI, while the below-frontier AI would be handled elsewhere.

We have this categorization:

• (1) Frontier AI.
• (2) Below Frontier AI.

What other ways might we divide up AI?

Some believe that it is better to consider the risks associated with AI (this has become both famous and infamous due to the EU AI Act; see my coverage at the link here). There is a claim that an AI doesn’t have to be a frontier AI to necessarily pose grave risks. The frontier AI is likely to have greater risks, but not exclusively so. If our attention is going to be on coping with AI risks, the belief is that rather than setting up an overseeing entity that is mandated to focus on frontier AI, it ought to be focused on risk levels instead.

Therefore, under that logic, we might have an entity that is directed at considering these risk levels:

• (1) Low-risk AI.
• (2) Medium-risk AI.
• (3) High-risk AI.

We could then declare that the entity is perhaps only to focus on high-risk AI, and allow the medium-risk and low-risk to be handled elsewhere. Or we could say that the entity encompasses medium-risk and high-risk, and leave the low-risk for our avenues.

Devising A New Entity

The Google proposal recommends that an independent entity be established. There is plenty of precedent for this in many other areas of societal domain-specific oversight. The proposal identifies several common examples.

Even that aspect of an independent entity carries controversy. Some ardently believe that a new independent entity is going to be a distraction. It will need to be created from scratch, and dedicated workers will need to be hired. The entity might start to become bloated. It could veer from its designated mission. The entity might fight for resources to survive, drifting from its main purpose. On and on, these open-ended horror stories are expressed.

What other options might there be?

The usual possibilities are these:

• (1) Assign an existing U.S. governmental agency to oversee all AI governance.
• (2) Establish a new U.S. governmental agency to oversee all AI governance.
• (3) Establish a new independent organization with governmental agency oversight.

The third one is the gist of the Google proposal, and the document provides its rationale for going that route. The other options are more heavily tilted toward the U.S. government having direct oversight of frontier AI. Some would say they are much more comfortable with the entity being a formal governmental organization. Others would decry that this is worse by far than an independent organization that would have arms-length governmental agency oversight.

Your choice is likely based on your perception of which can do a better job and how much the U.S. government should have a direct versus indirect role in overseeing frontier AI.

The Big Picture Viewpoint

Let’s consider the big picture positives of an independent nationwide entity that would oversee frontier AI and have a federal agency semblance of oversight:

• Creates a single overarching federally mandated focal point with a focus on frontier AI.
• Would consist of dedicated experts on AI, national security, economics, etc.
• Provides for the establishment of across-the-board consistent AI standards and safety guidelines.
• Improves AI national security coordination.
• Collects and analyzes AI critical incidents akin to aviation safety.
• Promotes international cooperation regarding frontier AI advances and releases.
• Reduces regulatory duplication within and between existing federal agencies and is hopefully nimbler than conventional governmental efforts.
• And so on.

Now that we have the upbeat side, let’s think about the tradeoffs and downsides:

• Determining what constitutes frontier AI is still murky and could leave out AI that ought to be encompassed within the scope of this entity.
• The below-frontier AI could still pose grave risks and is seemingly left to a sketchy and scattered range of oversight options.
• The independent entity might not move quickly enough to cope with the rapidly evolving pace of frontier AI and fall behind its crucial mission.
• The entity might foster a risk of regulatory capture, wherein AI makers get cozy, and the entity no longer sufficiently performs its independent role.
• If there are costs associated with dealing with the independent entity, this might aid the large AI makers and be a disadvantage or discouragement to startups and smaller AI players.
• The entity might go down the path of gradual mission creep and find other avenues for pursuing its interests, drifting from its core mission and splintering its attention.
• It might get into bitter fights with existing federal agencies, leading to extensive and exhaustive court cases to try to settle these disagreements.
• And so on.

I kept those lists to about seven key points each. Please know that there are a lot more upsides and downsides. That is to be expected and not a surprise. No matter which route we go, there are going to be weighty tradeoffs. There isn’t a rabbit in a hat that is simply waiting to be put on display.

The World We Are In

For those of you who are keenly interested in AI policy and AI governance, especially those who are AI ethicists and AI legal beagles, you ought to give a close read to the Google proposal. I suppose it is obvious to point out that Google has a big stake in these matters, and they are not on the sidelines. A smarmy cynic would say that whatever is proposed would be a sign of something that the tech bros want and, ergo, should be opposed. I don’t think we are going to make progress on these challenging matters if that’s the viewpoint that is going to be utilized.

We need to keep the dialogue going. Advances in AI are coming at a frenetic pace. A lot is on the line. What can we do to mitigate AI risks and yet retain and support AI advantages and benefits? This is the demonstrative question facing us now, and the answer will inexorably shape our future.

As the great statesman and philosopher Francis Bacon once remarked: “They are ill discoverers that think there is no land, when they can see nothing but sea.” Let’s keep our minds open – I’m optimistic that we can find a fruitful place to land.