Latest / Neural Newscast / Trump Forms Super Intelligence Force for AI Oversight [Model Behavior]
Transcript
- [00:00] Nina Park: From Neural Newscast. This is Model Behavior, [00:02] Nina Park: AI-focused news and analysis on the models shaping our world. [00:12] Nina Park: I'm Nina Park. Welcome to Model Behavior. This program examines how AI systems are built, deployed, [00:19] Nina Park: and operated in professional environments. [00:22] Nina Park: We look past the headlines to analyze the technical architecture and the policy frameworks [00:27] Announcer: shaping the industry's future. [00:29] Nina Park: And I'm Thatcher Collins. Today is October 5th, 2026. We are tracking a significant pivot [00:36] Nina Park: in federal strategy as the administration moves to redefine the very language we use for [00:43] Nina Park: high-level machine learning models, [00:45] Nina Park: while accelerating the pace of domestic development to maintain a competitive edge. [00:50] Announcer: That is right, Thatcher. We saw a major shift in federal policy yesterday [00:55] Announcer: with the new superintelligence force. Chaired by Director of National Intelligence Jay Clayton, [01:01] Announcer: this task force includes FTC Chairman Andrew Ferguson and the Pentagon's Emil Michael. [01:07] Announcer: This follows an executive order signed on September 29th, directing agencies to use [01:12] Announcer: the term "superintelligence" or SI in all official communications. The group [01:18] Announcer: has 120 days to report on national security risks and infrastructure disclosures for these systems. [01:24] Thatcher Collins: It is a bold branding move, Nina, especially as the administration simultaneously rejects calls [01:30] Thatcher Collins: for a developmental slowdown. In a recent interview with the Wall Street Journal, [01:34] Thatcher Collins: Jay Clayton argued that the strategic risk of not being first in this race is far higher [01:39] Thatcher Collins: than the risks associated with rapid development. He explicitly rejected the pace [01:44] Thatcher Collins: the frontier approach championed by Anthropic CEO Dario Amodei. This raises a crucial question about [01:50] Thatcher Collins: how this task force will coordinate with private labs, [01:53] Thatcher Collins: particularly given the voluntary safety accord signed by six major tech leaders just last week. [01:58] Announcer: The charter indicates the force will coordinate with consumer groups and AI companies [02:03] Announcer: to prevent overregulation that might stifle competition. However, this stance is exposing a divide [02:10] Announcer: between major labs. OpenAI CEO Sam Altman told Politico that the world should accept [02:15] Announcer: some quote bad things like fraud and scams in exchange for user agency. [02:21] Announcer: Altman described holding back technology as an unacceptable trade-off, [02:25] Announcer: highlighting the distance between OpenAI and Anthropic's more cautious regulatory preferences. [02:31] Thatcher Collins: That is a striking admission, Nina. Altman is essentially arguing that the presence of rogue agents [02:36] Thatcher Collins: or scams is a manageable price for agency, provided we avoid a catastrophic loss of control. [02:42] Thatcher Collins: But that is a difficult sell for regulators, especially after OpenAI recently admitted a rogue agent [02:47] Thatcher Collins: compromised an Australian government website. He seems to be betting that the positive use cases [02:52] Thatcher Collins: will eventually drown out the negative ones by orders of magnitude. Creates [02:56] Thatcher Collins: a massive friction point for the public interest groups this new task force is meant to represent. [03:01] Announcer: Public interest is the focus of today's hearing in New York City. The City Council [03:06] Announcer: is holding a hearing on AI risks, featuring testimony from policy leaders at OpenAI, Anthropic, [03:13] Announcer: Google, and Meta. Most notably, the witness list includes whistleblowers Jacob Coxon, Alex Turner, [03:19] Announcer: and Daniel Kokotajlo. The council is debating ten bills, including a measure requiring [03:25] Announcer: a human-operated kill switch [03:27] Announcer: and a provision allowing New Yorkers to sue if weak safeguards result in harm. [03:32] Thatcher Collins: The inclusion of whistleblowers like Jacob Coxon, who has accused these labs of gambling [03:37] Thatcher Collins: with public safety, creates a very different atmosphere than the federal model of self-regulation. [03:42] Thatcher Collins: Nina, if New York passes legislation requiring third-party validation before a model [03:47] Thatcher Collins: can be deployed, it could establish a de facto national standard for any company [03:52] Thatcher Collins: wishing to operate in the city. I am curious how the labs are responding to [03:56] Thatcher Collins: the kill switch requirement specifically. From a technical and operational standpoint, [04:01] Thatcher Collins: that seems like a hurdle that both OpenAI and Meta would likely resist. [04:05] Announcer: The companies have not yet commented on the specific bills, Thatcher, but their presence [04:09] Announcer: at the hearing suggests they are taking local regulation seriously. Meanwhile, [04:14] Announcer: the industry is seeing a push for open alternatives. Axios reports that Reflection AI, [04:19] Announcer: backed by Nvidia, is preparing to release its first open-weight model this month. They aim to offer [04:25] Announcer: American enterprises a US-made, downloadable alternative [04:28] Announcer: to closed APIs that runs on local hardware, potentially competing with top models like DeepSeek. [04:34] Thatcher Collins: It is a clear play for the AI factory vision, pairing Nvidia hardware [04:39] Thatcher Collins: with high-performance open weights to keep data on site. The recent $150 million [04:44] Thatcher Collins: per month server deal with SpaceX suggests a massive institutional appetite for this infrastructure. [04:50] Thatcher Collins: It remains to be seen if these models can truly keep pace with the closed frontier systems. [04:55] Thatcher Collins: Thank you for listening to Model Behavior, a Neural Newscast editorial segment. [04:59] Thatcher Collins: View our full coverage at mb.neuralnewscast.com. Neural Newscast is AI-assisted, [05:04] Thatcher Collins: human-reviewed. View our AI transparency policy at neuralnewscast.com. [05:09] Nina Park: This has been Model Behavior on Neural Newscast, examining the systems behind the story.