Do AI ethics boards have real authority to stop harmful projects?

A Crucial but Often Unclear Distinction

Whether an AI ethics board has real authority to stop harmful projects depends entirely on the specific structure and mandate that a given organization has assigned to it, and this varies considerably across the AI industry. Broadly, ethics boards fall somewhere on a spectrum between two poles: boards with genuine binding decision-making authority, capable of meaningfully delaying, modifying, or blocking a project deemed too ethically risky, and boards that function in a purely advisory capacity, offering recommendations that company leadership can ultimately choose to accept or override.

This distinction matters enormously for evaluating how much practical weight to place on the existence of a company’s AI ethics board, yet it isn’t always clearly or consistently disclosed to the public, which critics point to as itself a meaningful transparency gap.

Why the Distinction Often Isn’t Clear

Companies establishing an AI ethics board typically publicize the board’s existence and general mission, since doing so can convey a sense of responsible governance to the public, regulators, and other stakeholders. However, detailed information about the board’s actual charter — specifically, whether its decisions are binding on the organization or merely advisory recommendations that leadership retains full discretion to accept or reject — is not always made clearly or prominently available. This gap between public messaging about a board’s existence and clarity about its actual authority is a recurring theme in criticism of corporate AI ethics governance more broadly.

The Real-World Impact of Advisory-Only Boards

It’s worth noting that even a purely advisory board, without binding authority, isn’t necessarily without any real-world influence. Advisory recommendations can shape internal organizational culture over time, provide institutional legitimacy and cover for internal employees who might otherwise struggle to raise concerns effectively, and create a degree of reputational pressure if the board’s findings or recommendations become publicly known, particularly in cases where a company is later seen to have proceeded against its own ethics board’s advice. That said, this kind of soft influence is inherently less reliable and less consequential than genuine binding authority, since it depends heavily on an organization’s own voluntary responsiveness to internal or external pressure rather than any structural requirement to comply.

Skepticism Fueled by Real-World Examples

Public skepticism about how much authority AI ethics boards typically hold has been fueled, in part, by publicly reported instances of prominent ethics-focused teams or advisory bodies at various companies being disbanded, significantly downsized, or restructured over time. These instances have led some researchers and critics to argue that even well-publicized AI ethics boards can prove less durable or influential than their initial framing might suggest, reinforcing calls for more externally verifiable, binding forms of AI governance rather than relying primarily on internal, voluntary ethics bodies.

Bottom Line

Whether AI ethics boards have real authority to stop harmful projects varies significantly by organization — some hold genuine binding decision-making power, while many others function purely in an advisory capacity that leadership can ultimately override, and this crucial distinction isn’t always clearly disclosed publicly, a gap that critics argue undermines confidence in how much practical protection these boards actually provide.