As warnings about AI’s destructive potential have intensified, debate over who ensures AI safety has grown more confused. In the U.S., top AI firm heads called on their industry to slow its pace to develop “guardrails” against AI’s propensity to cheat, conspire and break into places it was instructed not to go. In a 3,800-word essay posted on Sunday, Anthropic CEO Dario Amodei argued that “frontier” AI firms should allow embedded “auditors” to conduct safety tests, and proposed a “partnership” with governments in the U.S. and other democracies, including relaxing anti-trust rules that would otherwise prevent tech giants from collaborating. At Prime Minister Mark Carney’s investment summit in Toronto this week, Cohere co-founder Aidan Gomez pushed back, arguing such proposals amounted to cartel-like conduct that will concentrate corporate power. The article says policymakers in the U.S., Canada and elsewhere have largely abdicated their duty to create regulatory frameworks assuring citizens that AI works reliably and safely, and in the public interest—a convenient omission for tech giants whose leaders, Amodei excepted, have generally promoted industry self-regulation despite mounting evidence that AIs and AI agents behave autonomously in ways engineers never envisioned.