Public transcript

What Did AI Leaders Just Agree To?

By: Special Competitive Studies Project
Length: 42 min
Published: 30 September 2026
Language: English
Text from: the audio
Transcribed: 1 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

This episode of the President's Tech Brief analyzes the 'morally binding' document signed by AI CEOs after meeting with the President, covering its four layers of control and open questions about implementation. Guest Brandon Daniels, CEO of Exeger and an attendee, describes the three-hour White House meeting and the consensus that the U.S. must win the AI race while embedding safety in code. The hosts also reflect on the U.S. withdrawal from Iraq, and a bipartisan 'Across the Aisle' segment discusses regulation, congressional action, and the new America.gov chatbot.

Key points

The President's meeting with AI CEOs lasted three hours and produced a document he calls morally binding, with four layers of control.
Brandon Daniels says everyone in the room agreed the most important thing is that the U.S. wins the AI race, while also embedding governance and safety directly into code.
Mark Zuckerberg and Jensen Huang circulated a set of superseding principles before the meeting, which was shared with Speaker Johnson and the attending CEOs.
The president proposed that third-party external auditors be overseen by a board that could include lawyers and policymakers, not just technologists.
Open questions include implementation deadlines, who the third-party auditors will be, whether reports will be public, and what happens when companies fail to comply.
Mike Kaiken and Josh Hodges debate whether this differs from prior voluntary commitments, with Josh pointing to FTC investigations of frontier labs and possible congressional action.
The hosts and guests discuss the launch of America.gov, a chatbot for government services, praising the initiative while flagging data readiness and cybersecurity risks.
The episode also marks the U.S. withdrawal from Iraq, with Martin and Ilbert sharing personal memories of serving there.

Questions it answers

00:00Show open

What is the President's Tech Brief?

It's a twice-weekly briefing on technology and national security that the SCSP team would put on the Resolute desk, tracking the tech competition with China through a tech scorecard.

The show frames the century's contest as fought in labs, code, and chips.
This episode analyzes the document signed by AI CEOs after their lunch with the President.

04:42The signed document

What are the open questions around the AI CEOs' signed document?

Martin calls the four layers of control a good starting point but wants to know how a failed test changes company operations, who the third-party auditors will be, implementation deadlines, and whether reports become public. He argues transparency builds trust.

The document has four layers of control the President expects companies to follow.
The FTC case against OpenAI and Anthropic shows enforcement is already in play.

08:06Leaving Iraq

How do the hosts reflect on the U.S. withdrawal from Iraq?

Martin and Ilbert share personal service memories from Ramadi and three tours including the 2007 surge. They note nearly 4,600 U.S. service members were lost and hope this withdrawal doesn't create a vacuum like 2011 when ISIS returned.

Martin served in Ramadi 16 years ago with the CIA.
Both would have preferred keeping a residual U.S. presence.

15:06Inside the White House meeting

What happened in the President's three-hour meeting with AI CEOs?

Brandon Daniels describes robust debate on superintelligence, safety, and infrastructure, with principles prepared beforehand by Mark Zuckerberg and Jensen. Everyone agreed the U.S. must win the race, and the group committed to governance in code, third-party auditors, and board oversight.

The President proposed auditor boards including lawyers and policymakers, not just technologists.
Daniels says implementation will happen within days, faster than regulation.
Companies will sign the document in the following days.

27:18Across the Aisle

How do Democratic and Republican strategists view the agreement?

Mike Kaiken and Josh Hodges both praise industry-government dialogue. Josh emphasizes verification and policing backstops, citing FTC reviews of frontier labs, while Mike urges waiting until after the election and pursuing enduring bipartisan congressional action.

Josh expects congressional action and cites FTC investigations of frontier labs.
Both see bipartisan opportunity in NDA provisions on AI.
DOJ, FBI, and the intelligence community are expected to play enforcement roles.

36:00America.gov chatbot

What do the guests think of the new America.gov chatbot?

They view it as a positive but overdue initiative, noting the U.S. lags countries like Estonia and Kazakhstan on unified government services. They stress data readiness and cybersecurity, since America is the largest cyber target in the world.

Government data mobilization is essential but not being done urgently.
Cybersecurity must move forward alongside the initiative.

40:59Closing

What's coming next from SCSP?

The hosts close the Thursday show and promote their second AI Plus Robotics Summit on October 14th, with registration at scsp.ai, returning Tuesday at noon.

Next show is Tuesday at noon across all platforms.

Notes

Overview

This episode of the President's Tech Brief examines the document signed by AI CEOs after a three-hour meeting with the President, which the President calls a morally binding agreement with four layers of control.

The Signed Document

Martin sees the four layers as a good starting point but asks what happens when a test indicates a failure, and how that changes how a company operates.
Unknowns include implementation deadlines, who the third-party auditors will be, and whether reports will be made public.
Martin argues transparency should be maximized to build trust, starting with an independent auditor everyone trusts.
The FTC case against OpenAI and Anthropic is cited as a development still in play.

Brandon Daniels on the White House Meeting

The meeting lasted three hours and featured robust debate encouraged by the President on superintelligence, safety, scalability, and infrastructure.
Mark Zuckerberg and Jensen put together superseding principles beforehand, shared with Speaker Johnson and the attending CEOs.
Each CEO had five to ten minutes to speak; motives of Nikesh, Haktan, and Dario were described as very different.
Everyone in the room agreed the most important thing is that the U.S. wins the race.
Daniels described governance in code, including a 'surge protector' that shuts off agents draining too much cost, compute, or cybersurface.
The President suggested third-party auditors be overseen by a board that could include lawyers and policymakers, not just technologists.
Daniels says companies will sign and implement the principles within days, faster than regulation could act.

Iraq Reflections

Martin served in Ramadi 16 years ago with the CIA; Ilbert served in Iraq three times, including with General Petraeus's staff during the 2007 surge.
Nearly 4,600 U.S. service members were lost; both hope the withdrawal does not create a vacuum as in 2011 when ISIS returned.

Across the Aisle

Mike Kaiken and Josh Hodges agree industry-government dialogue is positive; Josh stresses the need for verification and policing backstops.
Josh points to FTC reviews of frontier labs and expects congressional action; Mike urges waiting until after the election, 34 days away.
Both see bipartisan opportunity in NDA provisions to maintain the U.S. AI lead.

America.gov

The new government chatbot is praised as essential but boring-sounding; Estonia and Kazakhstan are cited as ahead on unified government apps.
Cybersecurity risk is flagged since America is the largest cyber target in the world.

Transcribed automatically from the audio. Summary and notes written by AI from the transcript, so check anything important against the recording.