Emil Michael Sold $50M in AI Rival Stocks While Setting Pentagon AI Policy
The top Pentagon official overseeing U.S. military artificial intelligence sold his stake in AI search engine Perplexity for between $5 million and $25 million in June 2026 — a transaction revealed by federal financial records on September 1 and not previously reported. The divestment comes four days after a federal judge permanently barred the Pentagon from enforcing a blacklist of Anthropic that Emil Michael helped engineer while holding between $2 million and $10 million in stock in Perplexity, one of Anthropic’s direct competitors.
The Perplexity sale completes a pattern: across three companies — xAI, Perplexity, and financial software developer Brex — Michael divested stock totaling as much as $55 million or more while simultaneously serving as the Pentagon’s chief technology officer, directing which AI firms gained access to the largest government AI deployment in American history. The realized gains are almost certainly legal. Federal ethics law almost certainly permits them. That gap between what the law prohibits and what public-interest standards would demand is the story federal officials, ethics experts, and Congress have yet to address.
Who Is Emil Michael
Michael, 53, is the Under Secretary of Defense for Research and Engineering — a position that also carries the title of Pentagon Chief Technology Officer. Confirmed by the Senate on May 20, 2025, he holds authority over research, development, and technology acquisition for the Defense Department, with artificial intelligence at the center of all of it. In practice, that makes him one of the most powerful individuals in the world determining which AI companies win multibillion-dollar government contracts and which face existential obstacles.
Before entering government service, Michael was chief business officer at Uber from 2014 to 2017, departing amid documented controversy. He subsequently accumulated a portfolio of investments in Silicon Valley companies across AI, cryptocurrency, and robotics — stakes that, at their disclosed ceiling values, were worth far more in government service than they were before he arrived.
The Perplexity Sale and What Preceded It
Michael’s relationship with Perplexity extended well beyond equity ownership. According to his 2025 financial disclosure filings, he received a personal loan from Perplexity of between $250,000 and $500,000 at a 4.57% interest rate and served on the company’s advisory board — resigning from that board upon entering federal service, while retaining the stock.
The Lever first reported Michael’s Perplexity holdings in March 2026, noting that he owned between $2 million and $10 million in the company’s stock at the time. At that point, Perplexity had secured a contract with the General Services Administration in November 2025, making its technology available government-wide — though no direct contractual relationship with the Pentagon has been publicly established.
What changed by June 2026 is that Michael sold those shares. The proceeds, per federal records reviewed by media outlets, fell somewhere between $5 million and $25 million — the broad ranges used in required financial disclosures mean the exact profit remains unknown. Perplexity has sought a market valuation of approximately $30 billion; the company was valued at roughly $20 billion as of early 2026.
A Three-Company Divestiture Ledger
The Perplexity sale does not stand alone. Earlier in 2026, Michael sold his investment in xAI — the company behind Elon Musk’s Grok model and a direct competitor to Anthropic — for between $5 million and $25 million. He had previously reported that stake as worth between $500,000 and $1 million, making the return an estimated gain of anywhere from 400% to 4,800%, depending on where within the disclosed ranges the actual figures fell.
The timing of the xAI sale drew particular scrutiny at the time. Michael received a divestiture certificate from the Office of Government Ethics on December 18, 2025, but did not execute the sale until January 9, 2026 — four days after the Pentagon announced a new xAI agreement and seven months after xAI’s initial $200 million Pentagon contract was announced. The Pentagon subsequently moved Grok onto classified military systems, work previously reserved for Anthropic’s Claude. Whether Michael participated in any decision directly affecting xAI’s government business has not been established by public records.
A third transaction rounds out the pattern. Michael also sold shares in Brex, a financial software developer acquired by Capital One in April 2026, for at least $5 million — a gain of at least 473% over his maximum disclosed holding of $750,000. A further detail: while Michael was still an investor in Brex, he appeared as a guest on a tech podcast called “Sourcery” in December 2025, in an episode titled “Misunderstood Power: Emil Michael on the Truth about DOW and Defense Tech” — a program sponsored by Brex. Brex has no documented Pentagon contracts.
Combined ceiling: xAI ($25 million) plus Perplexity ($25 million) plus Brex ($5 million minimum) equals a total stock divestiture of at least $55 million across the three companies during Michael’s tenure as the Defense Department’s chief technology officer.
What Federal Ethics Law Permits — and Doesn’t
The central legal question about Michael’s stock holdings and sales is not the one that sounds most dramatic. Reporters and ethics critics have repeatedly asked: did he violate federal law? The more accurate and more important question is different: what does federal law actually prohibit here — and why does it not prohibit the rest?
Federal conflict-of-interest law, codified at 18 U.S.C. § 208, prohibits a federal employee from participating “personally and substantially” in any “particular matter” in which the employee has a financial interest, if that matter could have a “direct and predictable effect” on that interest. The key term is “particular matter” — defined in the statute and Office of Government Ethics guidance as a specific proceeding, contract, claim, or determination affecting identifiable parties.
What § 208 does not prohibit is an official exercising general policy authority in ways that systematically favor or damage categories of companies, even when those officials hold financial stakes in the benefited companies. Determining that Anthropic constitutes a “supply chain risk” — a category designation affecting every defense contractor’s relationship with the company — is not the same legal thing as signing a specific Perplexity contract. Shaping a department-wide AI strategy that elevates xAI and OpenAI while marginalizing Claude is not a “particular matter” affecting an “identifiable party” in the statutory sense, even if it produces identical financial consequences for the companies involved.
Richard Painter, who served as the top White House ethics lawyer under President George W. Bush, offered a straightforward assessment: “He should have sold all interest in the company before he started working. That’s the way we would have done it back when I was working for president.” Painter’s standard — complete divestiture before service — is what prior administrations’ ethics cultures demanded. It is not what current law mandates.
The Pentagon’s response has been consistent and technically defensible: Michael and other defense officials “are in full compliance with ethics laws and regulations. Any claims otherwise are false.” The department described its ethics framework as “a rigorous, multi-layered system that includes financial disclosure reviews, divestitures where appropriate, and screening to prevent conflicts of interest.” What neither the Pentagon statement nor any public record has addressed is whether Michael was formally screened out of decisions involving xAI, Perplexity, or Brex — and what mechanism, if any, required or enforced such a screen.
The Anthropic Dispute as Context
The Perplexity and xAI holdings give the Anthropic blacklisting a context that was documented at the time but that the Perplexity sale now makes more concrete. In February 2026, Michael led efforts to designate Anthropic a “supply chain risk” under 10 U.S.C. § 3252, a statute previously reserved for companies with ties to foreign adversaries such as Huawei and ZTE. The designation followed Anthropic CEO Dario Amodei’s refusal to allow Claude to be used for fully autonomous weapons systems or mass domestic surveillance of American citizens.
At the time of the blacklisting, Michael held between $2 million and $10 million in Perplexity stock — in a company that is Anthropic’s direct competitor in AI-assisted search and enterprise AI. By designating Anthropic a supply-chain risk and directing every defense contractor to sever ties with the company within six months, the Pentagon’s action would predictably benefit Perplexity and other Anthropic competitors. Anthropic’s CFO warned in filings that the government’s actions could reduce Anthropic’s 2026 revenue by “multiple billions of dollars,” with documented disruption exceeding $180 million in affected negotiations.
On August 28, 2026, U.S. District Judge Rita F. Lin permanently barred the Pentagon from enforcing the designation, ruling in a 59-page opinion that the blacklisting violated Anthropic’s First Amendment right to free speech and its Fifth Amendment due process protections. Lin found the designation was driven by a desire to make a public example of Anthropic for publicly criticizing the government’s AI demands — not by any articulable national security threat. She noted the government’s risk-assessment memo was dated four days after the directives it was supposed to justify. The Trump administration is expected to appeal to the Ninth Circuit. A separate case under a different statute (41 U.S.C. § 4713, the Federal Acquisition Supply Chain Security Act) remains pending before the D.C. Circuit after oral argument on May 19, 2026.
Four days after Lin’s ruling, federal records disclosed that Michael had already cashed out his Perplexity holdings in June.
Senator Warren’s Pre-Confirmation Warning
The disclosure is not a surprise to at least one member of Congress. Senator Elizabeth Warren raised conflict-of-interest concerns in a March 2, 2025 letter to Michael before he was confirmed, writing that his “ties to technology companies that may seek contracts with the Department of Defense” warranted scrutiny — and specifically asking him to divest from holdings impacted by DoD’s contracts and investment priorities. She also noted a “history of inappropriate conduct,” a reference to the circumstances of his Uber departure.
Michael was confirmed anyway. Warren’s letter requested that he agree to recuse from matters involving former clients, divest from affected holdings, and commit to a four-year post-employment lobbying ban. The response her office received has not been publicly disclosed. No Senate investigation into Michael’s conduct has been publicly announced as of publication.
A Framework Designed for a Different Era
The Michael disclosures reveal a mismatch between the ethics infrastructure Washington has and the ethics infrastructure the AI era requires. Federal conflict-of-interest law was designed for a government that moves in procurement cycles measured in months and years, where the difference between a general policy decision and a specific contract award is clear. In the AI industry, a single policy decision — a supply-chain designation, a strategy memo directing “any lawful use” language into all contracts, a deployment authorization — can shift billions of dollars of competitive advantage within days.
The Project on Government Oversight documented a broader pattern through its Pentagon Revolving Door Database: the movement of senior defense officials between government and the private defense sector creates structural opportunities for financial benefit that existing cooling-off periods and specific-matter prohibitions were not designed to address. The Silicon Valley-to-Pentagon pipeline adds a new dimension — executives who accumulated equity stakes in pre-IPO companies whose valuations may increase significantly once those companies secure government contracts, creating a financial incentive structure that existing disclosure requirements document but do not prevent.
The Pentagon’s AI procurement budget is growing rapidly. Defense officials have described AI as central to American military superiority, and officials like Michael sit at the intersection of that spending, the companies competing for it, and the competitive intelligence that flows from operating at the top of the national security apparatus. Whether the ethics framework adequate to an older, slower industrial procurement environment is adequate to this moment is a question Congress, oversight bodies, and the public are pressing with increasing urgency. The answer, so far, is that the framework has not changed.
Frequently Asked Questions
Did Emil Michael break federal conflict-of-interest law by holding and selling AI stocks while directing Pentagon AI policy?
Almost certainly not — and that is the more important part of the story. Federal ethics law (18 U.S.C. § 208) prohibits federal employees from participating in specific “particular matters” — defined as contracts, rulings, or proceedings affecting identifiable parties — in which they have a financial interest. General policy decisions, including category designations (such as labeling a company a “supply chain risk”), broad strategy memos, and department-wide AI deployment authorizations, are not “particular matters” under the statute’s plain language. That structural gap means an official can hold millions in stock in companies that benefit directly from general policy decisions, as long as no specific contract or procurement action crosses their desk with that company’s name on it. The Pentagon’s statement that Michael is “in full compliance with ethics laws and regulations” is almost certainly accurate as a matter of law — but that accuracy does not mean the law is adequate.
How much did Michael make selling AI and tech stocks during his Pentagon tenure?
Federal financial disclosures use broad value ranges rather than precise figures, so the exact totals are unknown. The ceiling figures: xAI sale (January 2026): up to $25 million; Perplexity sale (June 2026): up to $25 million; Brex sale (April 2026 or after): at least $5 million. Combined ceiling: approximately $55 million. The xAI gain alone represented a return of between 400% and 4,800% over his previously disclosed holding value of $500,000 to $1 million. Brex represented a gain of at least 473% over his maximum disclosed holding.
What does the Anthropic ruling mean for other AI companies, and does Michael’s financial record affect it?
Judge Lin’s August 28, 2026 ruling established that the government cannot use 10 U.S.C. § 3252 to punish a domestic AI company for publicly stating what its technology will and will not do. Every AI lab holding government contracts and maintaining an acceptable-use policy — OpenAI, Google, Cohere, and others — now has constitutional protection for those policies under the California ruling. Michael’s financial record does not affect the legal outcome, but it shapes the historical record: the official who pressed hardest for Anthropic to abandon its guardrails held and subsequently sold stock worth up to $25 million in Anthropic’s direct competitor. Lin’s opinion did not need to address the conflict-of-interest question; she found the retaliation unconstitutional on its own terms. The D.C. Circuit FASCSA case remains pending.
What reform would actually close the gap in federal ethics law for senior AI officials?
The gap exists because the “particular matter” limitation in 18 U.S.C. § 208 was designed for a procurement environment in which general policy decisions and specific contract decisions were more easily distinguished. Closing the gap for senior officials with AI-sector portfolios would likely require one or more of the following: mandatory divestiture of all AI-sector holdings upon confirmation as a condition of service (not just recusal from specific matters), a broader definition of “particular matter” to include category-level designations and department-wide strategy decisions, or a prophylactic blind-trust requirement for senior DoD officials holding tech-sector equity. Congress has not advanced legislation specifically targeting this gap. Senator Warren’s request that Michael agree to four-year post-employment restrictions and full divestiture was made in her March 2025 pre-confirmation letter; whether he agreed to any such terms has not been publicly disclosed.
