A growing rift among artificial intelligence leaders over whether to slow development for safety reasons took center stage on Chicago’s Morning Answer, as host Dan Proft laid out competing views from industry figures before turning to a guest who dismissed the entire debate as corporate maneuvering.
The discussion came after a high-profile resignation from an Anthropic employee who cited fears of an AI-driven catastrophe, a move that drew support from prominent AI skeptics. Anthropic CEO Dario Amodei, appearing on CBS over the weekend, said the industry needs to proceed cautiously and argued that a coalition of democratically elected governments should share oversight of AI development, framing regulation as a way to give the public and elected officials a check on private companies. OpenAI CEO Sam Altman, in a separate podcast appearance, said his company is in no rush to pursue an IPO given current safety concerns, though he maintained the timing decision is about business readiness rather than outside pressure. Their comments stood in contrast to venture capitalist Marc Andreessen, who argued that open-source AI should be allowed to proliferate without regulatory barriers and that the United States should use its full private-sector and governmental power to secure global AI dominance over China.
Missy Cummings, professor and director of the Mason Autonomy and Robotics Center at George Mason University, rejected the safety-focused framing from Amodei and Altman outright, calling the current moment the biggest grift in the industry’s history.
She argued that AI companies are struggling financially and that their sudden interest in government partnership is really an attempt to secure public money without accepting public control. She said Altman’s reluctance to pursue an IPO has less to do with safety than with concerns that the technology cannot yet perform at the level investors would expect, which she said would hurt the stock price. Cummings pointed to a public rebuttal from investor David Sacks, who said AI companies do not need government intervention to justify slowing down if they genuinely believe their own unreleased models are too risky, and said she agreed with that assessment completely.
Cummings said the industry has repeatedly overstated what its models can actually do, citing recent coverage suggesting an AI system had solved a decades-old mathematical theorem when in fact the underlying proof had already been established by a human decades earlier, with the AI merely writing code to demonstrate it. She said she uses AI tools regularly and finds them useful, but that none of her robotics colleagues have managed to successfully connect a large language model to real-world robotic tasks outside of staged demonstrations, and that claims of imminent general or super intelligence are unfounded.
Asked about a recent MIT report examining how AI is reshaping campus life, Cummings said the findings mirror what she sees daily in her own classroom, including students offloading critical thinking to AI tools at the expense of developing problem-solving skills. The report found that a large share of MIT undergraduates use large language models daily and that many worry about their own overreliance on the technology. Cummings said she has restructured her own courses in response, basing grades entirely on in-person midterms, finals and class participation, with no books, notes, computers or phones allowed during exams. She said some students drop the class once they learn about the format, but that those who remain report a stronger grasp of the material, which she takes as evidence that educators need to rethink how they assess whether students are actually learning rather than relying on AI-assisted work.