In March, seven of the largest technology companies in the United States sat at the White House and signed the Ratepayer Protection Pledge. Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI committed to build, bring, or buy the power their data centers require, pay for the delivery infrastructure, and protect ordinary customers from the cost.
The objective is worthy. The public should not have to subsidize private computing demand. The commitment is also a useful test of how we talk about technology governance.
A pledge tells us what an organization says it will do. Governance tells us how anyone can know whether it did it.
Those are not the same thing.
A Promise Can Be Serious and Still Be Incomplete
It would be easy to dismiss every voluntary commitment as public relations. That would be unfair and unhelpful. Companies can move faster than legislatures. Technical teams often understand emerging risks before statutory language can be written. Voluntary standards can establish vocabulary, normalize safer practices, and create evidence that later rules can use.
The National Institute of Standards and Technology built its AI Risk Management Framework for voluntary use. Its value does not come from pretending voluntarism is enforcement. It comes from specifying work: govern, map, measure, and manage. It asks organizations to define responsibilities, document risks, test systems, monitor outcomes, and decide whether a system should proceed.
The White House pledge likewise contains concrete commitments. Signers say they will add power supply, cover infrastructure costs, accept separate rate structures, pay even when they do not use all contracted power, invest in local workforces, and coordinate with grid operators on resilience.
Those are more useful than a vague assurance that companies will act responsibly.
But specificity in a promise is still not proof of performance.
Governance Begins Where the Announcement Ends
For a commitment to become governance, the public needs answers to practical questions.
What counts as new or additive generation? Who verifies that a project would not have happened anyway? Which transmission, distribution, interconnection, reserve, and reliability costs belong to the data center? Can those costs migrate into ordinary rates years later? What happens if projected demand is cancelled after a utility has already built capacity? Who measures local employment promises? What data will be published, how often, and in what form?
Then come the harder questions.
Who has authority to investigate? Who can require a correction? What happens when a company misses a target? Can affected customers challenge the accounting? Are state regulators using compatible definitions, or can the same cost be treated differently across jurisdictions? Is evidence preserved long enough to compare a promise with actual bills, capital projects, and grid conditions?
Governance is not a press release followed by trust. It is a chain of authority, evidence, review, and consequence.
Evidence Must Be Designed In
The phrase “trust, but verify” is not enough if verification is added after decisions have already been made. Evidence has to be part of the operating design.
For a ratepayer pledge, that could include public project baselines, cost-allocation filings, independent engineering estimates, contracts that identify who bears cancellation risk, periodic reports using consistent metrics, and plain-language comparisons between projected and actual customer impacts. It should include incident reporting when assumptions fail, not only success stories when a project performs as planned.
The same principle applies to AI safety commitments.
An organization can promise red-team testing, model evaluation, privacy protection, provenance tools, or human oversight. The public still needs to know what was tested, against which threat model, by whom, with what limitations, and what changed because of the result. A summary can protect legitimate security and trade secrets while still establishing that a real process occurred.
Evidence does not require publishing every sensitive detail. It requires enough durable, reviewable information to support a conclusion.
Measurement Is Not the Same as Accountability
Metrics can create an illusion of control. A company can count tests, reports, meetings, or megawatts without proving that the underlying risk was reduced.
Good governance connects a measure to a decision.
If a safety evaluation identifies a severe capability, who decides whether release is delayed? If promised generation falls behind schedule, who prevents the cost from shifting to customers? If a model causes a pattern of harmful outcomes, who has authority to restrict, repair, or withdraw it? If an independent reviewer disagrees with the company, which conclusion controls?
The decisive question is not whether data exists. It is whether the data can change what happens next.
This is why NIST treats governance as a cross-cutting function rather than a final compliance step. Roles, risk tolerance, monitoring, and decommissioning have to be established before a system produces a crisis. Measurement without authority can describe a failure beautifully while doing nothing to stop it.
Regulators Still Matter
The Ratepayer Protection Pledge acknowledges a reality that technology companies cannot solve alone. Electricity prices and infrastructure obligations are shaped through utilities, grid operators, state commissions, contracts, and public law. A company may promise to pay its way, but public institutions still decide whether the arrangement is lawful, fair, and reliable.
That is not an argument for reflexive bureaucracy. It is an argument for assigning the right authority to the right institution.
Companies should build competent internal controls. Independent experts should test technical claims. Standards bodies should define interoperable methods. Regulators should protect rights, markets, safety, and public infrastructure. Courts should remain available when duties are disputed. Legislatures should define boundaries and consequences where voluntary action cannot protect the public interest.
No single layer is sufficient.
Self-regulation is strongest when it produces evidence that outside institutions can examine. Public regulation is strongest when it is technically informed, proportionate to risk, and capable of adapting without abandoning due process.
The Public Should Be Able to See the Difference
There is a moral difference between a company that makes a measurable commitment and one that offers only soothing language. We should recognize that difference.
There is also a governance difference between a measurable commitment and an enforceable duty. We should not erase that difference either.
Voluntary commitments can be a beginning. They can establish direction, accelerate useful work, and demonstrate that stronger standards are feasible. They can also delay binding action, fragment accountability, or create a reputation benefit without a matching public benefit.
The way to tell is evidence.
Show the baseline. Show the method. Show the result. Show who verified it. Show what happens when the result is not good enough.
That standard is not hostile to innovation. It is how serious institutions earn durable trust.
A promise may be made in a room full of cameras. Governance begins after the cameras leave.
Editorial Source Notes
* The White House, “Fact Sheet: President Donald J. Trump Advances Energy Affordability with the Ratepayer Protection Pledge”: https://www.whitehouse.gov/fact-sheets/2026/03/fact-sheet-president-donald-j-trump-advances-energy-affordability-with-the-ratepayer-protection-pledge/
* The White House, “Ratepayer Protection Pledge”: https://www.whitehouse.gov/ratepayer-protection-pledge/
* The White House, “President Trump Secures Historic Commitment to Keep Electricity Costs Down Amid Data Center Boom”: https://www.whitehouse.gov/releases/2026/03/president-trump-secures-historic-commitment-to-keep-electricity-costs-down-amid-data-center-boom/
* National Institute of Standards and Technology, “Artificial Intelligence Risk Management Framework”: https://www.nist.gov/itl/ai-risk-management-framework
* NIST AI Risk Management Framework Core: https://airc.nist.gov/airmf-resources/airmf/5-sec-core/
Editorial accuracy note:
The Ratepayer Protection Pledge is described as a public commitment. The article does not characterize it as a statute, regulation, or independently verified guarantee. It distinguishes the stated commitments from the evidence, enforcement, and public-authority mechanisms needed to establish durable governance.