Digital Colliers Daily Briefing — May 14, 2026
Three storylines define the AI industry today, and each cuts at a different layer of the stack: governance, privacy infrastructure, and commercial distribution. In a San Francisco courtroom, Sam Altman's personal investment portfolio and credibility became Exhibit A in a trial that could rewrite OpenAI's corporate structure. In Menlo Park, WhatsApp made the first serious attempt to reconcile AI inference with end-to-end encryption at consumer scale. And in expense-report data quietly compiled by a fintech, Anthropic crossed OpenAI in paid business customers for the first time — the same day it launched a product aimed squarely at America's 36 million small businesses.
1. Altman Concedes $2B in OpenAI-Adjacent Holdings as Credibility Becomes the Trial's Central Question
What happened. In testimony Tuesday before Judge Yvonne Gonzalez Rogers, Sam Altman acknowledged holding more than $2 billion in stakes in companies that do business with OpenAI, including a $1.7 billion position in fusion startup Helion Energy, according to a Reuters filing summary surfaced by Techmeme. Musk attorney Steven Molo walked Altman through prior accusations of dishonesty from former board members Helen Toner and Tasha McCauley, co-founder Ilya Sutskever, and Musk himself — and pressed Altman on his 2023 Senate testimony, in which he told Sen. John Kennedy he had "no equity in OpenAI" without disclosing economic exposure via a Y Combinator fund. Asked Tuesday if he could be trusted, Altman replied, "I believe I am an honest and trustworthy businessperson." Separately, testimony confirmed Microsoft has now spent more than $100 billion on its OpenAI partnership, per Wired's coverage from the courtroom.
Why it matters. The case is functionally a referendum on whether OpenAI's nonprofit board exercises real control over the for-profit — the legal hinge on which the company's restructuring depends. As TechCrunch noted, the November 2023 "blip" firing is being relitigated not for its own sake but as evidence that Altman's influence eclipses the board's. Bret Taylor testified Altman has been "forthright," and Satya Nadella called the firing "amateur city," but Taylor also conceded the rehiring was driven by the prospect of mass employee departure — not a clean exoneration. The $2B in adjacent holdings sharpens the conflict-of-interest narrative Musk's team has pursued from the start.
Who is affected. OpenAI's governance structure, Microsoft's $100B+ investment, and the precedent for how nonprofit-controlled AI labs convert to for-profit entities. A finding for Musk could force structural concessions; even a defense win leaves Altman publicly diminished going into a likely capital raise.
What to watch next. Closing arguments and the penalty phase, which Gonzalez Rogers will hear next week. Ars Technica reports Musk endured three days on the stand to Altman's roughly four hours — the contrast in courtroom posture may matter less than what the judge concludes about board control.
Sources:
- Musk v. Altman filing: Sam Altman holds $2B+ in stakes in companies that have done business with OpenAI, including a $1.7B stake in fusion startup Helion Energy (Kenrick Cai/Reuters) — Techmeme
- Who trusts Sam Altman? — TechCrunch AI
- Altman forced to confront claims at OpenAI trial that he's a prolific liar — Ars Technica
- Everyone at the Musk v. Altman Trial Is Using Fancy Butt Cushions — Wired
- Sam Altman's Private Texts Exposed In Trial — YouTube · Matt Wolfe
- Microsoft doesn't want any of this — The Verge AI
2. WhatsApp Ships Incognito Chat, Putting AI Inference Inside a Trusted Execution Environment
What happened. Meta announced Incognito Chat, a Meta AI chat mode inside WhatsApp built on the company's Private Processing infrastructure, which routes inference through hardware-isolated Trusted Execution Environments such that Meta itself cannot read inputs or outputs. The feature is text-only at launch, ephemeral by default, and runs on Muse Spark — the model Meta released last month. Per Wired, WhatsApp can see that an account used the feature but nothing about its content. Mark Zuckerberg, writing in a post cited by The Verge, called it "the first major AI product where there is no log of your conversations stored on servers." Meta is also rolling out "Side Chat with Meta AI," which lets users query Meta AI privately about an ongoing group chat without leaking that context to the model provider or other participants. Rollout to WhatsApp and the standalone Meta AI app will occur over the coming months.
Why it matters. Most "incognito" modes from OpenAI, Anthropic, and others suppress training and history but still allow the provider to see prompts and responses in transit and at inference. Incognito Chat is the first consumer-scale attempt to extend the end-to-end encryption guarantee to AI inference itself. Johns Hopkins cryptographer Matt Green, who advised on Private Processing, told Wired he has "confidence" the system delivers on its claim — while reiterating the standard caveat that any cloud secure-enclave system becomes a high-value attack target. WhatsApp head Will Cathcart's framing is telling: "Incognito Chat is kind of like we're running a giant phone for AI and we don't have the passcode."
Who is affected. WhatsApp's roughly 3 billion users, many of whom will encounter a frontier-class AI chatbot for the first time through this surface. Competitors — OpenAI, Anthropic, Google — now face pressure to match the architectural claim, not just the marketing one. Privacy-first players like DuckDuckGo and Proton, which TechCrunch notes have built similar offerings at smaller scale, get validation but lose differentiation. Worth flagging: Meta simultaneously removed opt-in end-to-end encryption from Instagram DMs this week, a contradiction Wired called out directly.
What to watch next. Third-party audit results on Private Processing, the timeline for image and voice support, and whether OpenAI or Anthropic respond with comparable enclave-based inference. Reuters reporting last month — cited by TechCrunch — that AI chat logs could be used in litigation gives the privacy pitch real legal teeth.
Sources:
3. Anthropic Passes OpenAI on Paying Business Customers, Targets 36M Small Businesses
What happened. Ramp's monthly AI Index, drawn from expense data across more than 50,000 client companies, shows 34.4% paying for Anthropic services versus 32.3% paying for OpenAI — the first time Anthropic has held the top spot, according to TechCrunch. A year ago, Anthropic sat at 9%. On the same day, Anthropic launched Claude for Small Business, a toggle inside its Claude Cowork agent platform that ships with 15 ready-to-run workflows and connectors to QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365. A ten-city promotional tour starts today in Chicago, with free half-day workshops for 100 local business leaders per stop and a one-month Claude Max subscription for attendees.
Why it matters. Ramp economist Ara Kharazian told TechCrunch that Anthropic's edge began in technical verticals — finance, tech, professional services — and is now spreading. OpenRouter's leaderboard, which samples a different user pool, last placed OpenAI above Anthropic in December 2025, supporting the broader trend. The small-business push is strategically significant because, as Anthropic notes, SMBs account for 44% of U.S. GDP and roughly half of private-sector employment, yet have lagged enterprise AI adoption. OpenAI shipped ChatGPT Business in late 2023; Anthropic is arriving later but with a deeper integration stack and a public-benefit-corporation framing that bundles training, CDFI partnerships, and a Workday Foundation solopreneur accelerator.
Who is affected. OpenAI most directly — Latent Space notes Anthropic's lead arrives just as OpenAI launched an enterprise switching promo offering two months of free Codex for migrating customers. The same Latent Space analysis flags a complicating signal: Anthropic this week capped subscription-subsidized programmatic usage, giving paid plans a fixed monthly API credit instead of effectively unlimited harness access. Power users including Theo and Jeremy Howard called it a rug pull; Kharazian himself was "skeptical about whether this advantage will last." The Ramp data also lands weeks before Anthropic's likely October IPO.
What to watch next. Whether the programmatic-usage repricing erodes developer-tool share to Codex, the conversion rate on the SMB tour, and the next Ramp Index reading to see if Anthropic's lead widens or proves cyclical.
Sources:
- Anthropic now has more business customers than OpenAI, according to Ramp data — TechCrunch AI
- Anthropic courts a new kind of customer: small business owners — TechCrunch AI
- [HN · 249↑] Claude for Small Business — Hacker News
- [AINews] Codex Rises, Claude Meters Programmatic Usage — Latent Space
Three stories, one pattern: the AI industry's competitive frontier is shifting from model quality to the institutional questions around it. OpenAI is being asked in court whether its governance can be trusted; Meta is being asked by the market whether its inference architecture can be; and Anthropic is showing that distribution discipline — technical verticals first, then SMBs, with training and integrations bundled — can flip a market that looked locked. Model leaderboards still matter, but May 14 is a reminder that trust, privacy, and channel strategy are now the variables that move share.