OpenAI Pushes AI Shields, Buys Talk Show Amid Leadership Scrutiny

OpenAI is simultaneously touting rapid technical progress, spending millions on media influence, and lobbying for legal protections that would shield it from responsibility if its systems contribute to mass casualties or billion-dollar disasters. The moves come as the company prepares for an initial public offering and its chief executive, Sam Altman, continues to reshape the organization after a period of notable internal turbulence.

On Thursday, OpenAI’s chief scientist, Jakub Pachocki, told the “Unsupervised Learning” podcast that recent gains in coding, mathematics, and physics suggest the company is nearing its goal of creating AI systems that perform like research interns. Pachocki said the key metric is how long a model can operate with minimal human supervision. OpenAI has set 2026 as the target for an “AI research intern” and 2028 for a fully autonomous researcher. The comments reflect a consistent theme at the San Francisco company: that artificial intelligence is advancing faster than many outsiders appreciate and that the economic rewards for staying ahead will be enormous.

Yet the same week brought reminders that OpenAI is also devoting considerable energy to managing its image and legal exposure. The company confirmed its purchase of “The Best Possible News,” a Silicon Valley streaming talk show, for an amount reported in the millions. Slate described the acquisition as either a sophisticated public-relations strategy or another example of free-spending by technology executives who now treat media properties like accessories. With an IPO on the horizon, the purchase gives OpenAI a direct channel to shape narratives about its technology among the very audiences most likely to adopt it.

At the same time, OpenAI threw its support behind Illinois Senate Bill 3444, which would grant broad liability protection to developers of so-called frontier models. The bill defines these models as those trained with more than $100 million in computing costs, a threshold that captures OpenAI’s latest systems as well as those from Google, Anthropic, Meta, and xAI. Under the legislation, companies would be shielded from lawsuits over “critical harms,” including those that kill or seriously injure 100 or more people or cause at least $1 billion in property damage, provided the harms were not intentional or reckless and the firm has published safety, security, and transparency reports.

OpenAI spokesperson Jamie Radice said in a statement that the approach focuses on “what matters most: Reducing the risk of serious harm from the most advanced AI systems while still allowing this technology to get into the hands of the people and businesses—small and big—of Illinois.” The company also expressed hope that the bill would discourage a patchwork of state regulations and encourage national standards.

Several AI policy experts told Wired the measure goes further than previous industry-backed proposals. A similar skeptical note appeared in coverage from Crooks and Liars, which highlighted the apparent contradiction between OpenAI’s warnings about existential risks and its desire to limit financial consequences if those risks materialize. The bill lists scenarios long discussed in AI safety circles: malicious actors using models to design chemical, biological, radiological, or nuclear weapons, or systems autonomously committing acts that would be criminal if done by humans.

This legislative push occurs against a backdrop of leadership questions that have dogged OpenAI since late 2023. A recent New Yorker profile, discussed at length on The Verge’s podcast, revisited Altman’s brief ouster as CEO and his swift return, after which he moved to reorganize the nonprofit-company hybrid structure and consolidate authority. Hosts David Pierce and Nilay Patel noted that Altman’s style resembles that of a conventional Silicon Valley operator more than a cautious steward of humanity-altering technology. Whether such a personality is suited to managing risks that could, in theory, involve mass death remains a live debate inside and outside the company.

The convergence of these stories illustrates a tension at the heart of today’s AI industry. Companies like OpenAI argue that rapid capability gains, such as those Pachocki described, will generate trillions in economic value and solve pressing scientific problems. They present themselves as responsible actors publishing safety reports and supporting measured legislation. Yet the request for liability carve-outs that apply only after $100 million has been spent on training suggests an assumption that the largest players should operate under different rules than smaller competitors or traditional software firms.

Economists have long observed that when the potential costs of failure are socialized while the gains remain private, incentives tilt toward speed over caution. OpenAI’s simultaneous pursuit of media influence, technical bragging rights, and legal immunity may reflect rational business strategy in an environment of intense competition with other well-funded labs. Whether that strategy aligns with the broader public interest, particularly if models begin to act with the independence executives now project, is a question state lawmakers in Illinois and beyond will soon confront.

The company’s trajectory also raises practical concerns about knowledge and accountability. Publishing reports is one thing; accurately anticipating every pathway by which a highly capable system could be misused or behave unexpectedly is another. History offers numerous examples of complex technologies whose risks became apparent only after widespread deployment. OpenAI’s preference for uniform national standards over varied state experimentation is understandable from a compliance standpoint, yet it risks locking in rules written before the full contours of the technology are known.

For now, OpenAI continues to recruit top technical talent, spend aggressively on both research and public perception, and lobby for boundaries around its legal exposure. Its executives sound confident that autonomous research interns are within reach. The question observers are left with is whether the company’s external behavior matches the gravity of the capabilities it claims to be developing. In an industry where fortunes can be made by moving first, the temptation to write one’s own liability rules is strong. Lawmakers tempted to accommodate that desire would do well to consider the long-term incentives such accommodations create.