Blog

Ensuring Public Trust: Moving from Self-Regulation to Co-Regulation

By David A. Hoffman // October 13, 2026

On September 29, 2026, President Trump and executives from Google, Meta, OpenAI, XAI, Nvidia and Anthropic signed the White House Accord on Super Intelligence (Accord). The Accord is a commitment that the signing companies will implement the following four layers of controls on their development of artificial intelligence (AI) technology:

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

These four layers of controls will be a good first step. They are necessary, but not sufficient. The accord further says that the companies “will meet regularly to establish standards and best practices to improve the safety of their systems.” Sam Altman recently said “(t)he world should trust that we are going to do the right thing because it’s the right thing and we feel the magnitude of this.” But trust is earned, or in this case has not yet been earned.

Whether it is Meta’s track record for stealing women’s reproductive health information, the harm they have caused to the mental health of children, OpenAI’s using other people’s intellectual property to train its models, or the recent evidence of models escaping testing sandboxes, several of these companies have demonstrated they do not yet deserve our trust. Fortunately, we have other options. Instead of just relying on the Accord, it could serve as the foundation for a co-regulation, rather than the proposed self-regulation model, giving these companies an opportunity to earn our trust.

I spent over two decades working in cybersecurity, privacy and data governance at Intel Corporation. During that time, we advocated for companies setting up processes where they could demonstrate that they develop technology and manage data responsibly, ultimately deserving of trust in the marketplace. I participated on non-profit boards and industry association convenings to develop a model that can work in this situation. I have recently published that model and titled it the Berlin Proposal, inspired by John F. Kennedy’s famous speech calling for a defense of freedom. 

This co-regulation model focuses on the creation of an independent verification organization (IVO) that sits outside of government. It recognizes that funding and responsible governance will be critical for the IVO. The IVO would have dual roles to develop risk management best practices, and to review accountability filings from the companies. Those filings would also subject the company and its executives to enforcement by the U.S. Federal Trade Commission and the State Attorneys General. That enforcement would include criminal prosecution for those executives who fraudulently sign the accountability representations. This model would provide a basis for these companies to demonstrate they deserve our trust. AI holds the promise to advance scientific exploration and help address some of the world’s greatest challenges. That scale of potential benefit is worth the investment in a co-regulatory model to govern it properly.

Here are the four essential elements of the proposal:

  1. The IVO - Creation of a new Non-Profit Independent Verification Organization (IVO)
    • Funded by mandatory annual contributions by the companies
    • The IVO Board should be elected by State Attorneys General (AGs)
  2. Published Best Practices - The IVO should publish annually updated AI Development and Implementation Best Practices
  3. Annual Filings - Companies would be required to file “Accountability Review” statements to the IVO and the U.S. Federal Trade Commission (FTC) and the AGs
    • Signed by the lab Board Chair and CEO
    • In the filing the companies voluntarily subject themselves to the oversight of the U.S. Federal Trade Commission (FTC) and the State Attorneys General 
    • Fraudulent signing of the filing would subject the executives to criminal prosecutions (similar to those under False Statements Act and False Claims Act)
    • Robust whistleblower protections should be put in place
  4. Substantive Organizational Accountability Requirements -The Accountability Review requirements should be based on long standing data governance best practices. These elements are based on those from the Center for Information Policy Leadership’s 2010 paper, Demonstrating and Measuring Accountability, of which I participated in the drafting process:

 

  1. Policies: Existence of binding and enforceable written platform safety policies and procedures that reflect applicable laws, regulations and industry standards. Organizations should develop, implement and publish safety policies (including privacy and cybersecurity) informed by appropriate external criteria found in law, regulation, or industry best practices, and designed to provide the individual with effective safety protections.
  2. Executive Oversight: Internal executive oversight and responsibility for data privacy and protection. Executive oversight will require the creation of a safety leader supported by appropriate resources and personnel, and responsible for reporting to organization leadership. Commitment by top management should include appropriate reporting and oversight of the organization’s safety program. Top management should empower and require senior-level executives to develop and implement the organization’s programs, policies and practices. 
  3. Staffing and Delegation: Allocation of resources to ensure that the organization’s safety program is appropriately staffed by adequately trained personnel. While recognizing the need to work within economic and resource constraints, accountable organizations should have in place sufficient staff to ensure the success of their safety program. Such staff should receive adequate training, both as they assume their role in the safety program and as that program evolves to address new developments in the organization’s business model, data collection practices and technologies, and offerings to consumers.
  4. Education and Awareness: Existence of up-to-date education and awareness programs to keep employees and on-site contractors aware of safety obligations. Organizations should provide the necessary briefings, information and education for their personnel to keep them apprised of current and emerging requirements. Such education should involve keeping employees aware of new safety issues that may affect the performance of their job, and sensitive to the importance of safety to individuals and to the success and reputation of the organization.
  5. Ongoing Risk Assessment and Mitigation: Implementation of a process to assist the organization in understanding the risks to safety raised by new products, services, technologies and business models, and to mitigate those risks. To be accountable, organizations must assess the safety risks raised by their products and practices as they are developed, implemented and evolve, and as their data requirements change. In response to the findings of those assessments, organizations must take measures to mitigate those risks.
    1. Safety Impact Assessments are one important risk assessment and mitigation tool. A Safety Impact Assessment is carried out as part of the process for determining whether to collect data, deploy a new technology or data-driven business model, or use or manage data in a particular way. It is also important when making decisions about how best to secure data. It involves close examination of each new application or process, an evaluation of its attendant risks, and a determination of the steps that must be taken to ensure that the organization meets the requirements of applicable law, regulation and the organization’s safety promises. To be accountable for its risk assessment and mitigation practices, organizations also should be able to demonstrate the nature of their risk analysis. The organization must show the rigor of the criteria against which analyses are carried out, and the suitability of those criteria to the nature of technology or service. Further, the organization should be able to demonstrate how decisions are made and steps are taken to mitigate risk. The organization must also demonstrate that the decisions it takes to respond to identified risks are appropriate and effective.
  6. Program Risk Assessment Oversight and Validation: Periodic review of the totality of the accountability program to determine whether modification is necessary. An accountable organization should periodically review its accountability program to ensure that it continues to meet the needs of the organization by supporting sound safety decisions. To encourage transparency, the results of that program review should be available to those persons or organizations external to the reviewing group tasked with program oversight. The method by which this information is derived and reviewed must be both appropriately rigorous and cost-effective for both organizations and regulators. The results of these assessment measures and/or audits should be reported to the appropriate personnel within the organization, and when necessary, corrective action should be taken.
  7. Event Management and Complaint Handling: Procedures for responding to inquiries, complaints and cybersecurity incidents. An accountable organization should implement a well-designed, reliable procedure for addressing safety problems when they arise. The procedure should include a formal complaint portion to address safety concerns of individuals, and potential or actual failures.
  8. Internal Enforcement: Internal enforcement of the organization’s policies and discipline for non-compliance. Accountable organizations should have in place policies and procedures for enforcement of its safety policies. Personnel who disregard those policies should be subject to sanctions, including dismissal.
  9. Redress: The method by which an organization provides remedies for those whose safety has been put at risk. Accountable organizations should establish redress mechanisms whereby individuals may have their complaints heard and resolved. The redress mechanism should be readily and easily accessible by individual, and address complaints efficiently and effectively. 

*I did not use AI in the writing of this blog, but I would not have been opposed to doing so.