Recent reports indicate that President Donald Trump, House Speaker Mike Johnson, and various technology chief executives are slated to convene for discussions regarding artificial intelligence (AI) on September 29.
This upcoming gathering arrives amid heightened momentum in the discourse surrounding AI safety. While prominent AI firms have sounded alarms regarding the potential hazards of the technology, alternative voices within the industry dispute the necessity of decelerating development.
Anthropic’s Call Meets a Cybersecurity Skeptic
Earlier this month, Jacob Coxon, a researcher at Anthropic, stepped down after claiming that both Anthropic and OpenAI were failing to act responsibly. Meanwhile, his peer Evan Hubinger estimated the probability of AI causing human extinction within ten years to exceed 10%.
Soon afterward, Dario Amodei, CEO of Anthropic, released an essay advocating for a reduction in the pace at which AI capabilities advance. This appeal received support from Sam Altman of OpenAI and Elon Musk of SpaceX.
Conversely, Nikesh Arora, CEO of Palo Alto Networks, expressed skepticism regarding the underlying intent. Via X, he characterized the advocacy for a slowdown as a covert business tactic. In a subsequent interview with CNBC, he suggested that these laboratories were seeking regulatory sympathy and an exemption from liability.
Arora now views such pacing as impractical, asserting that specific developers will persist in pushing frontier boundaries.
“I think more than likely that some people will jump the gun, and you’ll have still people developing at the frontier, which means we shouldn’t try and stop the frontiers because they’re the most responsible people,” Arora said.
Additionally, Arora rejected the 10% extinction probability assessment as exceptionally improbable.
Are AI Labs Actually Pumping the Brakes?
Amodei clarified in his essay that pacing would not entail pausing training activities, and the supporting labs have continued to deploy new models since the announcement.
Anthropic rolled out Claude Opus 5.5 on September 22, describing it as their initial model release following their call for a slower frontier. The firm noted that external assessment bodies, such as METR, examined the model prior to its debut.
On that same date, OpenAI introduced GPT-6 Sol and Luna. These offerings came on the heels of their flagship release, GPT-6 Astra, which arrived on September 4.
Google likewise maintained its release schedule, issuing Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15. Furthermore, Koray Kavukcuoglu of DeepMind disclosed that Gemini 4 has advanced to the post-training phase.
As these systems become available to the public, the Trump administration has commented on accountability for AI safety. On Monday, Treasury Secretary Scott Bessent asserted that AI creators ought to assume responsibility rather than anticipate federal liability protection.
This position offers little alignment with the liability shields Arora suspected the labs desired. To date, neither a complete attendee roster nor a comprehensive agenda for the September 29 meeting has emerged, leaving it uncertain whether the talks will focus on oversight, industry protections, national competitiveness, or a combination thereof.
Subscribe to our YouTube channel to watch leaders and journalists provide expert insights