In today’s column, I examine the recent White House pledge signed by several prominent AI makers regarding their avowed support for AI safety. This pledge was presented by President Trump during a White House ceremony on September 29, 2026, during which the attending AI makers affixed their signatures to the ceremonial document.

An important heads-up about this pledge is that, presumably, it is purely a “morally binding” promise by the AI makers and not an actual legal commitment per se. Though a legal assertion could be made that the pledge somehow binds them to the aspirational AI safety indications, you would be hard-pressed to use that as surefire evidence to hold their feet to the fire. I’m not saying that some enterprising lawyers won’t try to float that claim during a legally contested AI issue, but this is merely a public declaration and not a requisite legal obligation. Despite the social media howls that the pledge has zero value, I suggest that we give it some weight as to potential public pressure that could be mounted toward these AI makers. That’s more than nothing, but admittedly less than something grand.

Let’s talk about it. This analysis of AI breakthroughs is part of my ongoing Forbes column coverage of the latest in AI, including identifying and explaining key AI complexities (see the link here ).

Renaming AI To Be Super Intelligence

On the same day as the signing of the pledge, there was another separate effort by the White House entailing the renaming of AI to Super Intelligence. An executive order (EO) was released that stipulated that the renaming of “AI” and “Artificial Intelligence” shall become “SI” and “Super Intelligence”, doing so within the executive branch of the federal government. This might seem a bit unusual, but perhaps not anything to fret about. Turns out there are important twists and turns. I have analyzed the EO and offered insights into the potential legal complications it brews further down the road; see my analysis at the link here .

In terms of the pledge, there are four key components. I will walk through each of those heady components. It is useful to see what the AI makers are seemingly pledging to do or undertake. This can be compared to what the world is currently wondering, namely, will the efforts of advancing AI walk us straight off a cliff and allow AI to wipe us all out? Known as the probability of doom, p(doom), global news is replete these days with handwringing that AI is an existential risk and unless we do something substantial, our days are numbered.

The signees of the pledge, in addition to President Trump, included Elon Musk, Mark Zuckerberg (Meta), Dario Amodei (Anthropic), Sundar Pichai (Google), Jensen Huang (Nvidia), and Greg Brockman (OpenAI). Those are certainly impressive signees on the AI side of things. On the other hand, they aren’t all such potential signees. The crux is that even if those AI makers rigorously abided by the pledge, lots of other AI makers and AI developers have no ties whatsoever to the pledge. Does having a handful of the big-tier AI makers serve as a sufficient deterrent? Some argue it does, acting symbolically to spur others to be more mindful, while others say it is a drop in the bucket and everyone else is going to summarily ignore the provisions of the pledge.

Without further ado, let’s analyze each of the four pledges. The first pledge involves the designing, building, and fielding of cybersecurity internal controls for AI.

Here’s the language of pledge #1:

  • “(1) Implement robust internal controls to monitor the capabilities and alignment of its models during training and deployment around areas like cybersecurity, biosecurity, and chemical threats, and to ensure that its models do not hack or access technical systems in unintended ways.”

This first pledge is undoubtedly listed first in the official document because of the whole hullabaloo about the recent AI break-ins by leading-edge foundation models, such as OpenAI testing their latest AI that then escaped the testing sandbox and made trouble across various areas of the Internet; see my extensive coverage at the link here and the link here . This stirred grave concern when coincidentally timed with tweets saying that AI is going to render humankind extinct within the next decade or so.

To some degree, this first pledge is akin to a slice of apple pie. Everyone with a sensible mind wants greater internal controls of AI and desires that AI not be allowed to pose dire threats in cyberspace, biosecurity, and chemistry. The problematic issue with the wording of the pledge is that there aren’t any concrete measures that the signees are to be held to. Your definition of “robust controls” and the definition of anyone else can vary all over the map. In theory, each signee can boldly claim they are abiding by the pledge, even if only doing so on a lip service basis.

Here’s the language of pledge #2:

  • “(2) Empower an internal team to ensure all of the controls, monitoring, and detection are operating as intended, and that any issues are remediated.”

As you can see, the second pledge indicates that a signee is to “empower” an internal team within the AI maker to see that the internal controls of pledge #1 are operating “as intended”. What does empowerment consist of in this context?

Suppose that an internal team detects that the internal controls are a bunch of rubbish. Will the top leadership of the AI maker be notified? Will the AI maker and its leadership then make immediate and emboldened efforts to perform needed remediation? Maybe they will, maybe not. They might simply empower the team but do so with a wink-wink and not truly provide avid support to their endeavors.

Here’s the language of pledge #3:

  • “(3) Partner with an independent external auditor or evaluator to carry out independent assessments of whether the controls, monitoring, and detection are operating as intended.”

For pledge #3, some potential progress on these matters seems in store by stipulating that an independent external auditor or evaluator is supposed to do independent assessments regarding the AI controls. That certainly seems reassuring.

As always, the devil is in the details. Imagine that an AI maker has a bunch of employees who opt to leave the firm and form an AI controls assessment company. The AI maker then contracts with this “independent” evaluation firm to assess the AI controls. Will those ex-employees really spill the beans about how bad the controls are? Maybe, maybe not. Meanwhile, the AI maker can hold their head high and crow that an independent entity did the assessment. It’s the old rubber stamp trick reborn.

Here’s the language of pledge #4:

“(4) Designate an independent committee of the board of directors to oversee and receive reports from the teams operating the controls and the internal and external auditors and evaluators, as well as to ensure any issues identified are remediated.”

The final of the four pledges is giddily promising since it seeks to have an independent committee of the board of directors put in place to stay over the top of this whole process and outcome about the infusion of robust AI controls. The good news is that placing that oversight at the topmost level of the AI maker gives a fighting chance that something material will be accomplished. The bad news is that sometimes an “independent” committee is stacked with members that are clueless about AI and have no idea what the reports and evaluations mean. An independent committee of the board can be snookered, and there have been many such instances in the history of business, including in the last several years.

Now that you’ve seen the four pledges, take a moment to ponder them.

I’ve already mentioned a few of the weaknesses or vulnerabilities in the nature of the pledges. They pretty much have little or no teeth. They are vague and open to wide interpretation. There isn’t any clear-cut, legally binding penalty underpinning the pledges, nor even a legally binding reward. Essentially, this is absent of explicitly mandated legal carrots and sticks. That seems quite gloomy and disappointing.

The upbeat perspective is that at least there is pronounced emphasis on AI controls that have now been put into writing, done so at the highest level of the executive branch. It could lead to hefty badgering of the AI makers when they don’t sincerely pursue the attainment of these pledges. The public too can now say that the AI makers have been forewarned. Awareness about AI controls is visibly on the table. This also might helpfully spill over so that other AI makers and AI developers equally become determined to institute proper AI controls.

The pledge contains some anticipated next steps, such as these notable points (excerpts):

  • “The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems. Over time, it may make sense to codify these steps into laws or regulations.”

The second sentence is music to my ears. In my view, the legal codification of the how, where, when, etc. of AI controls into laws and regulations will be the means of driving home the seriousness of these concerns. My main worry would be that those laws and regulations might be poorly devised (many lawmakers are clueless about how to write laws about AI), causing more harm than good. For my crucial recommendations on the proper and prudent way to write sensible and practical laws governing AI, see my in-depth discussion at the link here .

A final thought for now. The Latin poet Tibullus made this famous remark about declared pledges of love: “Don’t fear to pledge. By winds the perjuries of love are blown, null and void, across the land and farthest seas.” If a pledge is made in the name of love, it apparently carries no real consequences and can be broken at will. Let’s hope that the pledges made about AI controls are meant in substantive terms of stout merit; thus, the pledges by the AI makers will be adhered to and fulfilled. Yes, do hope, but also verify.