Latest / Neural Newscast / Attorney Jay Edelson Slams OpenAI and Anthropic Heads [Model Behavior]
Transcript
- [00:00] Announcer: From Neural Newscast. This is Model Behavior, [00:02] Announcer: AI-focused news and analysis on the models shaping our world. [00:12] Nina Park: I'm Nina Park. Welcome to Model Behavior, a program where we examine [00:16] Nina Park: how artificial intelligence systems are built, deployed, and operated in [00:20] Nina Park: real-world professional environments. Today is September 16th, 2026, and we are looking at [00:26] Nina Park: a sharp escalation in the rhetoric surrounding AI safety and corporate leadership. Specifically, [00:32] Nina Park: we are analyzing the increasing friction between the industry's public messaging [00:37] Nina Park: and the actual risks communicated by internal researchers and legal experts. [00:41] Announcer: It is a critical shift, Nina. [00:43] Thatcher Collins: This centers on Jay Edelson, a prominent class action attorney who is currently leading [00:48] Thatcher Collins: major lawsuits against both OpenAI and Anthropic. In a recent interview with [00:53] Announcer: News Nation, Edelson. [00:54] Thatcher Collins: Edelson stated quite bluntly that the general public cannot trust anything coming from [00:58] Thatcher Collins: OpenAI CEO Sam Altman or Anthropic CEO Dario Amodei. It is a direct challenge [01:04] Thatcher Collins: to the personal credibility of the two most visible figures in the frontier model space today. [01:10] Nina Park: Edelson's argument is not based on general corporate skepticism, Thatcher. [01:14] Nina Park: He specifically called out what he perceives as an abdication of professional responsibility [01:19] Nina Park: across the industry. During his appearance on News Nation with Leland Vittert, [01:23] Nina Park: he pointed to a fundamental paradox: these leaders often argue that their products are [01:28] Nina Park: insanely dangerous, yet they continue to push them into the market at a rapid pace. [01:33] Nina Park: He suggested that if they truly believed [01:35] Nina Park: an extinction-level event was a realistic outcome, they would halt development immediately. [01:40] Thatcher Collins: You are right, and there is a significant scientific dimension to this that we should not ignore. [01:45] Thatcher Collins: Edelson is drawing on warnings from actual researchers within these organizations. [01:50] Thatcher Collins: He cited Jacob Coxon, a former researcher at both Anthropic and OpenAI, [01:54] Thatcher Collins: who recently warned that the very people building these systems earnestly believe they could cause [01:58] Thatcher Collins: massive harm by the end of the decade. [02:01] Thatcher Collins: This sentiment has been echoed by Evan Hubinger, who leads alignment science at Anthropic. [02:05] Nina Park: That leads us to the six-sided die metaphor Edelson used in his interview. Facher. [02:10] Nina Park: He described a scenario where one side of a die represents a mass extinction event, [02:14] Nina Park: while the other five sides represent progress or profit. He argued that Sam Altman [02:19] Nina Park: and other leaders are effectively willing to roll that die for the sake of rapid development. [02:24] Nina Park: It is a heavy accusation that challenges [02:27] Nina Park: the narrative that these executives are being responsible simply by talking about the risks. [02:32] Thatcher Collins: The reporting suggests a pattern of leadership admitting to a lack of total control [02:37] Thatcher Collins: while still pursuing superintelligence. Edelson argues they are proceeding [02:42] Thatcher Collins: because of a long-term profit motive. Interestingly, Nina Edelson did not lump every tech leader [02:48] Thatcher Collins: into this category. He pointed to Elon Musk and XAI as a counterexample, [02:53] Thatcher Collins: specifically citing [02:54] Thatcher Collins: the lawsuit Musk filed to force OpenAI back toward its original nonprofit, safety-first mission. [03:00] SPEAKER_03: It is a notable alignment between a trial lawyer and Elon Musk, both calling for a slowdown based on [03:06] SPEAKER_03: internal safety warnings. We have seen Sam Altman and Dario Amodei make public calls for regulation, [03:13] SPEAKER_03: but Edelson's core point is that these calls for caution are essentially performative [03:18] SPEAKER_03: if the underlying products remain active and accessible. If the researchers' own stated fears [03:23] SPEAKER_03: are accurate, [03:24] SPEAKER_03: then simply asking for government oversight while accelerating deployment seems contradictory. [03:29] Thatcher Collins: Exactly, Nina. The friction here is between the apocalyptic marketing used by AI leaders [03:35] Thatcher Collins: and the actual business practices they follow. If a product is genuinely viewed as having [03:40] Thatcher Collins: a non-zero chance of causing human extinction, typical safety protocols in other industries, [03:46] Thatcher Collins: such as aviation or pharmaceuticals, would dictate an immediate global halt. [03:51] Thatcher Collins: Edelson is highlighting that this standard [03:52] Thatcher Collins: has not been applied to AI development, and that is where the public trust begins to break down. [03:58] Nina Park: It raises the question of whether the danger narrative is being used to invite regulation [04:02] Nina Park: that favors incumbents by raising the barrier to entry, or if it is a genuine concern [04:07] Nina Park: being sidelined for commercial gain. Regardless of the motive, the lawsuits led by Edelson's firm [04:13] Nina Park: will likely force internal communications and safety data into the public record through discovery. [04:18] Nina Park: This could finally provide the clarity [04:20] Nina Park: we need regarding what these researchers actually know about the limits of model alignment. [04:25] Thatcher Collins: It is a debate that is quickly moving from the theoretical space of the laboratory [04:31] Thatcher Collins: into the high-stakes environment of the courtroom. As these cases progress, [04:36] Thatcher Collins: the transparency [04:37] Thatcher Collins: of the frontier model developers will be put to its ultimate test. I am Thatcher Collins, [04:43] Nina Park: and I'm Nina Park. Thank you for listening to Model Behavior, a Neural Newscast editorial segment. [04:50] Nina Park: You can find more of our deep dives into the AI industry at mb.neuralnewscast.com. Neural Newscast [04:58] Nina Park: is AI-assisted, human-reviewed. View our AI transparency policy at neuralnewscast.com. [05:05] Announcer: This has been Model Behavior on Neural Newscast, examining the systems behind the story.