Analysis
Giorgos Tsetis, cofounder of the supplement company Nutrafol and an active investor in AI companies including Anthropic, xAI, Positron AI and Lila, argues in a Fortune op-ed that AI executives face too little personal legal exposure for the harms their companies can cause -- and that this has to change before, not after, something goes seriously wrong.
Tsetis draws the comparison from his own experience: at Nutrafol, a company he says became 'multi-billion-dollar,' he felt personally legally responsible for product safety. He argues corporate fines don't replicate that pressure at AI labs, calling them 'just another line item' that amounts to 'a fraction of their turnover and valuations.' His framing is blunt: 'Responsibility that stops at the corporate entity isn't responsibility. It's an expense.'
“In late September, leaders of several major AI labs signed a White House accord covering internal risk controls, independent audits and board-level safety oversight.”
The stakes he cites are not abstract. Tsetis points to potential AI-driven harms including an individual suicide, large-scale cyberattacks and geopolitical destabilization as the kind of downside that justifies extending liability to named individuals, not just balance sheets. He treats social media as the cautionary precedent: an industry whose harms compounded for years under an accountability structure that never reached past the corporate entity.
He's arguing against a backdrop that already exists. In late September, leaders of several major AI labs signed a White House accord covering internal risk controls, independent audits and board-level safety oversight. Tsetis calls these steps 'welcome' but says they're voluntary -- exactly the gap he wants closed. He doesn't specify penalties or draw the line between civil and criminal liability, saying those thresholds belong to 'legislators, lawyers, technologists and ethicists.' His actual proposal is narrower than the op-ed's framing suggests: 'a named senior executive should be personally responsible for meeting' whatever standards get set.
Room for disagreement: a personal-liability regime for AI executives could just as easily push talent and headquarters toward jurisdictions with weaker rules, the same dynamic that's shaped crypto regulation for a decade. It could also chill exactly the kind of frontier research Tsetis's own portfolio depends on, since the executives most exposed to new liability would be the ones running the labs doing the riskiest work. Tsetis doesn't address either risk directly, and the thresholds he's calling for -- what counts as 'meeting' a safety standard, and when a harm is foreseeable enough to trigger liability -- are the hard part no accord or op-ed has actually solved yet.