Mozilla's Mark Surman on 3 Ways CEOs Can Build Trust in AI in 2026

Stephanie Mehta's May 18 2026 interview with Mozilla Foundation president Mark Surman is short on celebrity and long on something the AI conversation has been missing for two years. A working theory of trust. Surman has spent the last decade running the nonprofit arm of the organization behind Firefox, and his pitch to CEOs is not the usual "be transparent" platitude. He gave Fast Company two concrete moves, gestured at a third behind the paywall, and pointed at Harvard Business School professor Karim Lakhani's research as the intellectual scaffolding.

What makes the interview worth reading even if you never sit in a CEO chair is the audience Surman is actually addressing. He is talking to executives but the test cases are employees. The trust he wants built is internal first, external second. That inversion matters for anyone deciding which company to join, stay at, or leave in 2026.

What Surman's 3 Trust Moves Actually Are

The first recommendation Surman put on the record is "empower your team." On the surface this sounds like the same language every CEO has used since 2010. Surman tightened it into something testable. His exact words to Fast Company were "if you want to do right by your employees, have them be involved in how you reshape and rebuild the company." He followed with "give them ways to create and learn and have agency over how AI is used."

That is not a slogan. It is a specification. Three things have to be true for a company to pass Surman's test. Employees have to be in the room when AI plans are made. They have to be allowed to build with the tools rather than only consume outputs. And they have to retain agency, meaning the ability to say no, redirect, or shape the rollout. Companies that announce an "AI-first" memo on a Monday and ship it without internal consultation by Friday fail all three.

The second recommendation Surman named is "build the right guardrails." Fast Company's published excerpt does not include Surman's full elaboration on this point, but the framing is consistent with the trust gap that has been growing across enterprise AI deployments in 2026. Guardrails in his vocabulary are not just safety filters on a chatbot. They are the institutional rules that decide what AI can touch, who reviews its output, and what gets escalated to a human before it ships. Mozilla itself has spent years arguing that guardrails are an organizational design choice, not a vendor feature.

The third recommendation sits behind the Fast Company paywall in the published excerpt available at press time. Surman gestured at it but the verbatim text was not visible in the syndicated version. What is visible is the intellectual anchor. Surman cited Karim Lakhani's research at Harvard Business School. Lakhani has spent the last several years arguing that AI is not primarily a technology adoption problem but an organizational learning problem. The companies that win, in Lakhani's framing, are the ones that treat AI like a new colleague who needs to be trained, supervised, and integrated. Surman's third move almost certainly draws from that well.

Why Most CEOs Are Failing the Trust Test

Surman did not name companies that are getting it wrong. He did not have to. The pattern is by now familiar to anyone who has followed the wave of CEOs publicly walking back their own AI-first declarations, or watched recruiters drown under 300 applications per role while companies blame the volume on the same AI tools they deployed.

The default 2026 CEO playbook has three steps. Step one is announce an AI strategy on an earnings call. Step two is direct middle managers to "find efficiencies." Step three is post a LinkedIn thread about how the workforce of the future will be "smaller and more empowered." This is the precise sequence Surman's framework rejects. Employees are not consulted in step one. They are the variable in step two. And they read step three as the threat it usually is.

Microsoft's own 2026 research, which found that AI gains are flowing to leadership rather than to frontline workers, is the clinical version of the same diagnosis. The trust gap is not a perception problem. It is a distribution problem. CEOs are capturing the productivity and the upside narrative while pushing the displacement risk down the org chart. Surman's "empower your team" line is a direct rebuttal of that distribution.

The IBM 2026 CEO study reached a parallel conclusion from inside the C-suite itself. Most chief executives now privately concede that their AI bets carry job-loss risk, and a separate finding shows a growing share of CEOs fear losing their own jobs if those AI bets fail. The fear runs in both directions. What Surman is offering is a way out that does not require any CEO to pretend the risk is not there.

There is also a structural problem Surman did not raise but his framework illuminates. The 2026 AI acumen gap means many bosses cannot evaluate the systems they are buying. A CEO who cannot read a model card or stress test a vendor demo is going to lean on slide decks from the same vendors selling the tools. Guardrails built by people who do not understand the tool tend to be theater. Surman's "build the right guardrails" line implicitly requires CEOs to either learn or hire the literacy in.

How Workers Should Read Their CEO's AI Posture

For employees, the Surman framework is most useful as a diagnostic. You do not need to wait for a town hall to know whether your CEO is on the trust-building or trust-burning side of the line. Three signals tell you almost everything.

The first signal is whether the AI rollout was announced or co-designed. If you and your peers learned about the new tooling from a press release or a Slack post from the CHRO, you are working at a company that flunked Surman's first move. Co-designed rollouts look different. They involve pilot groups, internal working sessions, and revisions to the plan based on what employees say. They take longer. They also fail less often. The hidden cost of speed-over-quality AI deployments is one of the most underreported stories of 2026.

The second signal is whether your manager can explain the guardrails. Not the marketing version. The operational version. Which outputs require human review before they reach a customer. Which categories of decision the model is not allowed to make. What happens when the model is wrong. If your manager cannot answer those questions in plain language, the guardrails do not exist in any meaningful sense. They are slides. Companies treating guardrails seriously also tend to invest in structured AI team meetings and the three practices managers actually need.

The third signal is whether your job description has changed and whether you were part of the change. The 2026 reality is that nearly every knowledge-work role is being rewritten around AI assistance. Healthy companies are doing this in conversation with the people in the role. They are funding training, adjusting performance criteria, and updating the five skills that beat job titles inside AI workplaces. Unhealthy companies are simply raising the bar and waiting to see who falls off, a pattern that has produced the 2026 entry-level squeeze where AI raised the productivity bar for new hires without anyone telling them how to clear it.

For job seekers, the diagnostic runs in reverse. In interviews, ask the hiring manager three Surman-shaped questions. How were employees involved in your AI strategy. What are your guardrails and who enforces them. What does an employee do here if the AI is wrong. The quality of the answers will tell you whether the role is a real seat at a real table or a placeholder while the company waits to see what gets automated. The same logic applies inside the company. The first 90 days of any new job are still the window where trust gets built or burned, and the AI-shaped version of that window is now compressed.

There is one more reading of Surman worth holding onto. He is the president of Mozilla, an organization that survives on the trust of users who could switch browsers in 30 seconds. His instincts about trust are not abstract. They come from running a product that has to earn its place every day. CEOs whose companies enjoy switching costs measured in years and contracts have the luxury of getting trust wrong for a while. Mozilla does not. That is why Surman's framework reads as a working theory rather than a TED Talk. It was built under pressure, and it shows.

People Also Asked

Q: Who is Mark Surman and why does his AI advice matter?

A: Mark Surman is the president of the Mozilla Foundation, the nonprofit behind Firefox and a long-running voice on open-source technology and digital rights. His advice carries weight because Mozilla operates in a market where users can switch products in seconds, so the organization has spent decades learning how to earn and keep trust at scale. When he tells CEOs how to build AI trust, he is not theorizing. He is describing what has worked at Mozilla.

Q: What are the three ways Mark Surman says CEOs can build trust in AI?

A: Surman told Fast Company on May 18 2026 that CEOs need to empower their team and build the right guardrails. The third recommendation sits behind the publication's paywall in the published excerpt, but Surman tied his framework to Harvard Business School professor Karim Lakhani's research on AI as an organizational learning problem. The first two moves alone form a usable test for any CEO's AI posture.

Q: How can employees tell if their CEO is actually building AI trust or just talking about it?

A: Watch three signals. Whether the AI rollout was announced top-down or co-designed with employees. Whether your manager can explain the operational guardrails in plain language. And whether your job description changed in conversation with you or behind closed doors. If the answers are announced, no, and behind closed doors, your CEO is closer to the failing side of Surman's test than the passing side.

Ready to level up? Use Surman's three signals to read the AI posture of any company you are considering, then take that lens into your next interview or internal review. Join Metaintro to get the labor market intel you need to pick employers who treat AI as a team build rather than a top-down rewrite.