On TechCrunch’s Equity podcast, Anthony Ha, Kirsten Korosec, and Sean O’Kane discussed whether the artificial intelligence industry is ready to slow its development or move forward in a more controlled manner with a “pace the frontier” approach. The discussion began after Anthropic CEO Dario Amodei’s plan to “pace” the progress of frontier AI and similar remarks by OpenAI CEO Sam Altman. Nvidia CEO Jensen Huang, meanwhile, supported President Donald Trump’s view that the backlash against AI is a “hoax” and that regulation is unnecessary.
Sean O’Kane described Anthropic, OpenAI, and, to some extent, Elon Musk supporting the idea of slowing down as a significant change compared with their previous assessments. However, he argued that the plan Amodei outlined in his blog post did not contain sufficient detail about the reasons for the risks or what “slowing down” would mean in practice. Anthony Ha said that although the rapid consensus forming around the plan in the industry was encouraging, it also raised doubts about whether this was a real and large-scale step toward slowing down.
Kirsten Korosec said that this approach was of particular relevance to frontier AI laboratories and listed the main proposals as follows:
- Independent third-party evaluators working at companies such as Anthropic and OpenAI to monitor AI safety practices and incidents.
- Major AI companies in democratic countries coordinating on safety standards and limits.
- Establishing coordination at the international level.
Korosec said that the broad framework of these proposals was known, but the details remained unclear, and that there was also a possibility that the agreement could turn into the beginning of a cartel. She also stated that Nvidia CEO Jensen Huang had strongly opposed this approach.
According to O’Kane, Huang is an important intermediary between the industry and the Trump administration and has adopted a stance that supports the progress of AI companies. Korosec, meanwhile, argued that Huang’s position was linked to his interests, given that Nvidia stands to benefit greatly from AI advancing without restrictions. Ha also said that Trump’s and Huang’s interests aligned and that, although the proposals included safety measures, they did not explicitly require AI to advance more slowly than it does today.
The podcast also addressed whether existing regulations and the free market were sufficient for safety. O’Kane said that the federal government did not appear willing to enforce regulations and that consumer choice was not steering this market sufficiently. Arguing that companies sell more services to enterprise customers and are supported by investment capital, she said that, if a serious problem arose, customers could not be expected to easily switch from OpenAI’s Codex to Anthropic’s Claude Code.
Why it matters
The significance of the debate lies in the question of whether AI safety will depend solely on companies’ voluntary decisions or on independent oversight and common standards across countries. While the proposed framework is particularly relevant to labs developing frontier AI, the fact that these companies agree on the same rules could foster cooperation on safety, but it could also turn into a structure that limits competition. Nvidia’s approach, meanwhile, takes the debate beyond the boundaries of technical safety; it remains unclear to what extent the interests of companies and public authorities diverge regarding the pace of AI. The fundamental issue left unresolved, therefore, is how independent evaluation will work, which risks will be addressed at which thresholds, and whether existing regulations will actually bind companies in practice.