OpenAI Bought the Talk Show That Regularly Interviews Its Own CEO

01OpenAI Buys the Talk Show That Regularly Interviews Its Own CEO

Five days a week at 2 PM Pacific, TBPN goes live. For up to three hours, hosts interview AI executives, tech investors, and industry leaders on camera. Sam Altman has sat in the guest chair. So have executives from Meta, Microsoft, Palantir, and Andreessen Horowitz. Bloomberg and CNBC have treated the show as a peer outlet. By most measures, it was an independent media property covering the AI industry's most powerful players.

Now one of those players owns it.

OpenAI announced that it has acquired TBPN. The company says the deal will "accelerate global conversations around AI" and "support independent media." Both claims sit uneasily with the basic structure of the transaction: a company that is frequently the subject of coverage now controls the outlet doing that coverage.

The show's live format was its editorial asset. Executives came because the audience was engaged and the hosts gave them room to speak at length, outside the constraints of a cable news segment. Whether that access means the same thing when the booking call comes from a show owned by OpenAI is a question the company has not addressed.

OpenAI is not the first tech company to buy a media property. Jeff Bezos bought The Washington Post in 2013, and Salesforce's Marc Benioff purchased Time magazine in 2018. In both cases, the buyers argued that ownership and editorial independence could coexist. The Post has broadly maintained its autonomy, though questions about Amazon coverage have persisted. Benioff has kept a lighter hand. But neither buyer acquired an outlet where they personally appeared as a recurring on-air source.

That distinction matters. The Post covers thousands of subjects; Amazon is one. TBPN's beat is the AI industry. OpenAI is not a minor topic within that beat. It is, by funding and public attention, the central one. A talk show owned by its primary subject faces a conflict of interest that is structural, not hypothetical.

OpenAI frames the acquisition as expanding dialogue "with builders, businesses, and the broader tech community." In practice, it gives the company a daily live platform with an established audience and industry relationships built on editorial independence the show no longer has. OpenAI has not disclosed deal terms or announced changes to TBPN's editorial staff.

02Perplexity Sued Over 'Sham' Incognito Mode as Granola's 'Private' Notes Turn Out Public

A class-action lawsuit filed this week accuses Perplexity of operating an "Incognito Mode" that does nothing of the sort. The complaint alleges Perplexity, Google, and Meta shared millions of user conversations to increase advertising revenue. Users who toggled on the privacy feature believed their queries wouldn't be tracked or monetized. According to the suit, that belief was wrong.

Court filings claim Perplexity's incognito label borrowed user trust built by years of browser privacy modes, then failed to deliver equivalent protections. Google and Meta are named as co-defendants, with the three companies allegedly exchanging user chat data in arrangements that generated ad revenue. Conversation volumes cited in the complaint run into the millions.

Perplexity was not the only AI company whose privacy language parted from its product behavior this week.

Granola markets itself as an AI notepad for professionals in back-to-back meetings, with a clear promise: notes are "private by default." The Verge reported that any note is viewable by anyone who possesses its link. Users' notes also feed Granola's internal AI training pipeline unless they manually opt out. A product sold on privacy ships with sharing enabled and training opted in.

The two cases operate at different scales. Perplexity faces a legal challenge with named co-defendants and financial allegations. Granola has a design flaw that one settings change could fix. Both rely on the same borrowed vocabulary. "Incognito" and "private by default" carry expectations shaped by a decade of browser privacy features. Chrome's incognito mode doesn't share browsing data with advertisers. A "private by default" document in Google Docs isn't accessible via link without explicit permission.

AI companies adopted these terms without adopting the protections behind them. The lawsuit portrays Perplexity's incognito feature as a label on a data-sharing pipeline. Granola labeled a link-accessible, training-eligible note as private. One case heads to court. The other awaits a settings change.

03Google and OpenAI Roll Out Tiered AI Pricing in the Same Week

Three announcements landed within days of each other. Google released Gemma 4, its most capable open-weight model, free to download and deploy. The company also split the Gemini API into two inference tiers: Flex for cost-sensitive workloads, Priority for production reliability. OpenAI, separately, introduced pay-as-you-go pricing for Codex, opening the code-generation agent to ChatGPT Business and Enterprise teams without fixed commitments.

Each targets a different price point. Together they trace a pattern cloud computing established a decade ago: free tier to attract developers, usage-based billing to convert them, premium reserved capacity to lock them in. AWS, Azure, and GCP all followed this sequence. AI platforms are now running the same play, compressed into months instead of years.

The timing reflects a specific market condition. Model performance gaps between frontier labs have narrowed enough that switching costs are falling. When the top four or five models produce comparable output for most tasks, the developer's decision reduces to price, latency, and integration friction. Google's new Flex tier makes the tradeoff explicit: lower cost in exchange for variable latency. That only works as a selling point when the model itself is no longer the differentiator.

Gemma 4 adds a structural wrinkle. Google positioned it specifically for "agentic workflows," making it the first major open model to target multi-step agent orchestration by name. Agent frameworks today rely overwhelmingly on closed API calls, where every reasoning step incurs per-token cost. A capable open model running locally collapses inference cost to zero. For teams building autonomous coding or research agents, the economics shift from variable API expense to fixed compute.

OpenAI's Codex pricing responds to a related pressure from the enterprise side. Buyers have resisted per-seat licensing for tools their teams use sporadically. Pay-as-you-go removes that objection and signals that OpenAI sees Codex adoption as a volume game, not a premium upsell.

Both companies now compete on the axis that decided cloud market share: who offers the most granular pricing at each stage of developer adoption.