As warnings about advanced AI grow louder, a second question is emerging: whose expertise should carry the most weight when society evaluates the risks — the engineers building the systems, dedicated safety researchers, outside experts, policymakers, or all of them?
WHAT’S HAPPENING
Hugging Face CEO Clément Delangue questioned the attention being given to former Anthropic researcher Jacob Coxon, who recently resigned and publicly raised concerns about catastrophic AI risk.
Delangue compared asking Coxon about AI extinction risk to asking an air-conditioning technician about climate change.
He later clarified that he was not saying Coxon was necessarily wrong. His point was that Coxon primarily worked on AI pretraining rather than existential-risk research and that the discussion should include a broader range of expertise.
That includes AI developers, safety researchers, social scientists, policymakers, investors and others studying how advanced AI could affect society.
WHY IT MATTERS
Working inside a frontier AI company provides firsthand knowledge of how advanced systems are developing.
It does not automatically make someone an expert on every economic, social, security or long-term consequence of the technology.
The reverse is also true.
Experts outside AI labs may bring important perspectives, but they may not have the same direct exposure to what frontier systems can actually do.
The debate increasingly depends on combining both.
WHO BENEFITS
The public benefits when AI-risk discussions include multiple areas of expertise instead of relying on a single group or viewpoint.
AI developers can explain capabilities.
Safety researchers can test failure modes.
Social scientists can examine broader consequences.
Policymakers can evaluate public-interest concerns.
Independent researchers can challenge assumptions made inside the companies building the technology.
WHO LOSES
The discussion becomes weaker when individual opinions are presented as industry-wide conclusions simply because the person works — or previously worked — at a major AI company.
The same applies when legitimate concerns from people inside frontier labs are dismissed solely because others disagree with their conclusions.
Expertise matters.
So does evidence.
WHAT HAPPENS NEXT
As AI systems become more capable, disagreement over their risks is likely to continue.
The stronger debate will increasingly focus on what the systems can demonstrably do, what independent testing shows and which experts are qualified to evaluate each specific risk.
Straight Talk: building advanced AI gives someone an important seat at the table. It does not give any one person — inside or outside the industry — the entire table.
