Apple v. OpenAI Trade Secret Case

Apple v. OpenAI Trade Secret Case

Apple’s complaint in Case No. 5:26-cv-07078, filed July 10, 2026 in the Northern District of California, alleges conduct by Chang Liu that satisfies every element of criminal trade secret theft under 18 U.S.C. Section 1832, and the Department of Justice has convicted a defendant on an identical fact pattern before. The question is not whether the conduct described warrants a criminal referral. It does. The question is whether federal prosecutors will act on it, and what the answer means for every venture-backed AI hardware company that has built its technical team by recruiting engineers away from Apple, Google, and Nvidia. Why Does the Levandowski Conviction Matter for This Case? The Levandowski conviction is the governing precedent. In 2020, the Northern District of California convicted Anthony Levandowski of a single count of criminal trade secret theft under 18 U.S.C. Section 1832 for downloading approximately 14,000 files from Google’s self-driving car project before leaving to found a competitor later acquired by Uber. He received eighteen months in federal prison. The factual structure of the Apple complaint is the same: a senior technical employee, a mass download of confidential materials, a move to a direct competitor, and evidence of deliberate concealment. On several dimensions, Apple’s factual record is stronger than what existed in Levandowski at the time of indictment. Liu’s own communications, preserved on an Apple-issued device, establish knowledge of unauthorized access in his own words. The message quoted in the complaint describes his discovery that he could still access Apple’s network storage after his employment ended and his credentials should have been disabled. That is not circumstantial evidence. It is a written admission of the mental state Section 1832 requires: knowledge that the access was unauthorized. Liu also directed communications about the downloads to the LINE Messenger application, a choice the complaint frames as deliberate concealment, and he coached a current Apple employee on how to avoid Apple’s security team during the departure process. What Is Tang Yew Tan’s Specific Criminal Exposure Under Section 1832? Tang Yew Tan’s exposure is different in character but serious on its own terms. Twenty-four years at Apple is not a fact a prosecutor ignores. His alleged conduct centers on solicitation rather than direct computer intrusion: using internal Apple project codenames to probe job candidates for confidential information, directing candidates to physically remove Apple hardware from Apple premises, and distributing Apple’s internal departure security document to prospective hires before those individuals had given Apple any notice of their planned departure. Section 1832’s prohibition on obtaining trade secrets by fraud or deception covers structured deceptive interview practices. The dispositive open question is how Tan obtained the internal departure document after leaving Apple. If discovery answers that question unfavorably, his criminal exposure sharpens from a solicitation theory to a direct procurement theory, which is considerably easier to prosecute. What Is the Basis for OpenAI Corporate Criminal Liability? OpenAI’s corporate exposure turns on what its senior leadership directed and when. Apple’s complaint attributes the scheme not to rogue employees acting alone but to organizational conduct directed by the Chief Hardware Officer and reflected in structured interview protocols circulated across the organization. Corporate criminal liability under respondeat superior attaches when employees commit criminal acts within the scope of their employment and at least in part for the benefit of the corporation. The Yates Memorandum, issued by the Department of Justice in September 2015, requires federal prosecutors to identify culpable individuals before extending cooperation credit to the corporate entity. The content of OpenAI’s internal communications will determine whether the DOJ pursues individuals, the corporation, or both. Wire fraud under 18 U.S.C. Section 1343 reaches every electronic communication used in furtherance of a scheme to obtain property by fraud, and the conduct described in the complaint supplies those communications in quantity. What Does This Case Mean for Venture Capital Diligence in AI Hardware? For investors, the more immediate question is not the criminal docket. It is what this complaint signals about the category of risk embedded in AI hardware deals. OpenAI acquired io Products specifically to accelerate its hardware independence from incumbent chip suppliers. Apple’s complaint goes directly at that program. Trade secret litigation in the semiconductor sector routinely costs between $10 million and $50 million in legal fees before trial. The injunctive exposure is potentially more damaging than the monetary exposure: an injunction requiring the return or destruction of the alleged trade secret materials, or barring use of derived materials in product development, can halt a hardware program during the precise window when competitive timing matters most. The Levandowski matter ended Uber’s self-driving car program as originally conceived. A parallel criminal proceeding adds a layer of commercial disruption that civil litigation alone does not generate. Criminal defense retainers for individual defendants, Fifth Amendment privilege assertions in civil depositions, and plea agreements that produce adverse testimony against the corporate entity each create independent disruptions to product timelines and investor confidence. Any fund conducting diligence on an AI hardware company whose technical team includes material contributors from major incumbents must treat intellectual property provenance as a first-order question. The inquiry is whether those engineers’ work at the new company is susceptible to a credible allegation that it incorporates or derives from materials they were not authorized to take. What this case does not resolve is the broader question of what legitimate talent mobility looks like in a sector where the same engineers built the prior generation of hardware at the incumbents and are now being recruited to build the next generation at challengers. The statute does not prohibit engineers from changing employers or from applying skills and general knowledge at a new company. It prohibits taking specific, protectable, confidential materials without authorization and using them for a competitor’s benefit. Where exactly the line falls between general expertise and protected trade secret is a factual and legal question that will be contested in discovery and at trial. The Levandowski case produced a single conviction on a single count out of a far larger alleged scheme. What the Northern District of California does with Case No. 5:26-cv-07078 will set the next marker. Read my full analysis here: https://theinnovationattorney.com/blog/ This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theinnovationattorney.substack.com/subscribe [https://theinnovationattorney.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

Using DARPA's OTA as a Strategic Intellectual Property Tool

DARPA’s Other Transaction Authority under 10 U.S.C. Section 4022 lets defense technology startups negotiate IP ownership with the government and avoid the default government license the Federal Acquisition Regulation imposes. An Other Transaction Agreement for a prototype project sits entirely outside the Federal Acquisition Regulation and outside the statutory data rights scheme that automatically attaches Unlimited Rights, Government Purpose Rights, or Limited Rights to a standard defense contract. Because none of those default categories apply unless the parties write them into the agreement, a startup negotiates the intellectual property schedule from a blank page rather than from a government favorable template. The startups that benefit most treat that blank page as a drafting opportunity rather than paperwork to sign quickly so a program can start. I spent three years as a submarine officer aboard USS William H. Bates before spending three decades advising technology companies, and one pattern has not changed in that time: the government’s leverage over intellectual property comes from a license structure most founders never read until a term sheet is already signed. Other Transaction Authority exists because Congress recognized, and expanded the authority repeatedly since, that a young company building autonomous flight software or a novel sensor has no reason to accept the same data rights scheme built for a shipyard constructing a destroyer under a fixed price contract spanning a decade. What does 10 U.S.C. Section 4022 actually let the Pentagon do? 10 U.S.C. Section 4022 gives the Pentagon authority to enter into transactions other than contracts, grants, or cooperative agreements for prototype projects directly relevant to enhancing the mission effectiveness of military personnel or the equipment the armed forces use. The statute sits beside 10 U.S.C. Section 4021, which covers research transactions, but Section 4022 is the provision startups encounter most often because it governs the prototype stage, where a working demonstration, not a research paper, is the deliverable. To use the prototype authority at all, at least one of several conditions must be met: a nontraditional defense contractor or nonprofit research institution must participate to a significant extent, every significant participant must be a small business or nontraditional contractor, at least one third of the cost must come from non federal sources, or the agreement must have used competitive procedures with at least two competing proposals. Agreements between one hundred million and five hundred million dollars require a written determination from the head of the contracting activity or the relevant DARPA, Defense Innovation Unit, or Missile Defense Agency director, and anything above five hundred million dollars needs approval from the Under Secretary of Defense for Acquisition and Sustainment along with notice to Congress. How does an Other Transaction Agreement change who owns the resulting technology? An Other Transaction Agreement changes ownership outcomes because it removes the automatic license structure a Federal Acquisition Regulation contract imposes and replaces it with whatever the parties negotiate. Under a standard defense award, the government receives Unlimited Rights in technical data or software developed solely at government expense, Limited Rights in data developed solely at private expense that embodies a trade secret, and Government Purpose Rights, which convert to unlimited after a five year period, in anything developed with mixed funding. Because 10 U.S.C. Sections 3771 through 3775, the statutes that create this scheme, do not apply to a prototype Other Transaction Agreement unless the parties import them by reference, a startup can instead retain its background intellectual property outright, grant a license limited to the specific data delivered rather than everything created during performance, or shorten the government purpose period well below five years. The founders who capture this advantage negotiate the intellectual property schedule as a standalone exhibit rather than accepting whatever boilerplate a contracting officer attaches to the solicitation, since nothing in the statute requires favorable terms to be offered first. What happens when a prototype moves toward full production? 10 U.S.C. Section 4022 allows a follow on production contract to be awarded without further competition, but only if three conditions are met: competitive procedures were used to award the original prototype agreement, the prototype project was completed successfully, and the follow on production option was written into the prototype agreement from the start. The Government Accountability Office tested exactly this sequence in Oracle America, decided in 2018, sustaining a protest because the Army’s prototype agreement contained no follow on production provision and the awardee had not finished the prototype before the follow on award was made. The lesson for a startup negotiating its first Other Transaction Agreement is to insist that the follow on production path, along with the intellectual property terms that will govern it, is written into the original agreement rather than left for later, since a company that waits until the prototype succeeds to negotiate production terms has already given up most of its leverage. Can a competitor stop the government from using this authority? A competitor’s ability to challenge an Other Transaction Agreement is narrower than it would be for a Federal Acquisition Regulation contract, though it is not eliminated entirely. The Government Accountability Office does not review the substance of who should have won an award, but it will review whether the statutory conditions for using the authority were satisfied, which is exactly the theory that succeeded in Oracle America. Commentary on a 2025 Court of Federal Claims decision suggests that court views itself as an available forum for these disputes too, so a startup cannot know with certainty today which forum will hear a future challenge. A startup that documents compliance with the statutory conditions at the time of award, rather than assuming the Other Transaction label makes the agreement immune from challenge, protects both its funding timeline and its negotiated intellectual property terms. Two questions remain live for any startup building a contracting strategy around this authority. A startup cannot know today which forum, the Government Accountability Office or the Court of Federal Claims, will hear a future challenge to its award, since the Court of Federal Claims has only recently begun describing itself as available for these disputes. A startup also cannot assume the intellectual property template it receives this year already reflects the broader nontraditional contractor exemptions Congress adopted in the fiscal year 2026 National Defense Authorization Act, since the Pentagon has not yet published updated model schedules. Read my full analysis here: https://theinnovationattorney.com/blog This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theinnovationattorney.substack.com/subscribe [https://theinnovationattorney.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

Firestorm Labs Wins $30 Million Contract for 3D Printed FPV Drones

The Department of War awarded Firestorm Labs a $30 million APFIT contract on May 8, 2026, to deploy five containerized xCell drone microfactories and more than 200 Tempest unmanned aerial systems to the Indo-Pacific. The award was made under the Accelerate the Procurement and Fielding of Innovative Technologies program, a procurement mechanism established in Fiscal Year 2022 specifically for technologies ready to transition from development to operational units. Approximately $26 million has been obligated across five task orders and deliveries are underway, marking Firestorm’s first deployments outside the continental United States. This contract validates a production model first tested in Ukraine, where 3D-printed drones have become central to tactical operations and where the design cycle compresses from months to days. What Did the Squall Prove in Ukraine? Orqa FPV, the Croatian company that co-developed the Squall with Firestorm, has operated FPV drones in Ukraine since the early phase of the conflict. The Squall is a Group 1 FPV quadcopter: it reaches 80 mph, carries a 5.5-pound payload, flies for 42 minutes, and ranges 20 miles. Its airframe is produced by HP Multi Jet Fusion industrial 3D printing, not injection molding or traditional composites. All components are NDAA-compliant under the procurement restrictions established by Section 848 of the National Defense Authorization Act for Fiscal Year 2020, addressing Pentagon requirements and allied purchasing rules. Ukraine established specific operational facts for defense planners. Approximately 50 to 70 percent of Ukrainian drones are 3D-printed, according to multiple assessments. Drone designs iterate at a pace measured in days, driven by rapid adversary adaptation in electronic warfare and kinetic countermeasures. Fixed manufacturing facilities are targets: Ukrainian factories have been struck repeatedly, demonstrating that centralized production creates single points of failure in a contested environment. Firestorm CEO Dan Magy cited this experience in the company’s April 2026 Series B announcement, noting that the design cycle has compressed dramatically and that fixed production sites cannot be taken for granted. The Squall’s validation in Ukraine is the commercial foundation for the entire xCell model. A drone produced at the location of need, printed on-site rather than shipped from San Diego, changes how a commander manages attrition. The Squall demonstrated that HP Multi Jet Fusion airframes are not prototype curiosities. They are battlefield-ready platforms. What Is the xCell System and How Does It Work? The xCell is not a single 3D printer in a shipping container. It is a deployable industrial node built around HP Multi Jet Fusion printers, housed in two standard shipping containers, and designed to operate entirely off-grid. It can be airlifted inside a C-17 or C-130, sling-loaded beneath a CH-47 Chinook, or moved by sea. Firestorm holds a five-year global exclusive with HP for mobile deployment of Multi Jet Fusion technology, secured in July 2025. That exclusivity represents a commercially significant barrier: no competitor can replicate the manufacturing partnership in the near term. HP Multi Jet Fusion deposits binding and fusing agents across successive layers of nylon polymer powder, then uses a fusing lamp to selectively solidify the material into structural parts. The parts are strong, accurate, and repeatable at production scale. The weapons added to the finished drone are not 3D-printed; they are attached separately. The airframe is the product. That distinction matters for International Traffic in Arms Regulations export control analysis and for battlefield logistics planning. The Army used xCell to produce on-site replacement parts for a Bradley Fighting Vehicle, parts that would otherwise require months of conventional procurement. That application illustrates a capability extension beyond drone manufacturing: in a Pacific conflict where sea lines of communication are contested, the ability to print structural components for a range of systems adds value that the nominal drone mission does not capture. What Does the $30 Million APFIT Contract Actually Fund? APFIT does not fund research and development. It funds the procurement and deployment of technologies already proven and ready to field. Awards range from $10 million to $50 million for small businesses and non-traditional performers. The $30 million awarded to Firestorm on May 8, 2026, expandable to $50 million, is at the high end of that range. The package covers five xCell mobile manufacturing units, more than 200 Tempest drones, and operator training for an undisclosed Indo-Pacific customer. The Tempest, the airborne system paired with xCell in this contract, reaches approximately 400 miles, flies for six hours, and carries a 10-pound payload. It is configured for ISR and one-way attack missions at operational distances, not the short-range FPV tactical profile of the Squall. The combination of Tempest’s extended reach with xCell’s forward production capability creates a distributed aviation node that does not depend on rear-echelon resupply. Separately, Firestorm holds an Air Force contract with a $100 million ceiling, of which approximately $27 million has been obligated. Total company funding stands at $153 million following the $82 million Series B in April 2026, with investors including Lockheed Martin, In-Q-Tel, NEA, and Washington Harbour Partners. What Does This Mean for the Indo-Pacific? In the Indo-Pacific, U.S. and allied forces operate across vast distances, dispersed basing, exposed maritime corridors, and an adversary threat set built around long-range precision fires, anti-ship weapons, airfield attack, and anti-access and area-denial systems. Moving finished drones from the continental United States, Hawaii, Guam, or other fixed hubs into the first island chain may become slow, expensive, and contested in a crisis. By deploying xCell within the theater, the United States shifts part of its defense industrial capacity forward and complicates adversary planning built on severing logistics lines. Firestorm Chief Growth Officer Chad McCoy summarized the logic: if a blockade occurs, the machine does not stop. This framing reflects the Pentagon’s designation of contested logistics as one of only six national critical technology areas. The xCell model supports distributed maritime operations, expeditionary advanced base concepts, and dispersed air-ground teams by providing a production layer that moves with the force rather than waiting at a fixed installation. Three questions the industry is still working through: whether HP Multi Jet Fusion airframes can meet the structural requirements of one-way attack profiles at high delivery angles; how ITAR treats an xCell unit deployed under a foreign military sales or bilateral defense agreement in the Indo-Pacific; and whether the APFIT ceiling of $50 million is large enough to support the full-scale deployment Firestorm plans within two years. Read my full report here: https://theinnovationattorney.com/blog/ This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theinnovationattorney.substack.com/subscribe [https://theinnovationattorney.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

The Last Jet Ace

On May 10, 1972, Lieutenant Randy Cunningham and Radar Intercept Officer Lieutenant William Driscoll became the only United States Navy aces of the Vietnam War, shooting down three MiG 17s in a single mission over North Vietnam during Operation Linebacker. They were flying an F 4J Phantom II from Fighter Squadron 96 off USS Constellation, callsign Showtime 100. The engagement is taught at the Navy Fighter Weapons School to this day as the defining case study in Energy Maneuverability tactics applied under live fire conditions. What produced that outcome matters more than the outcome itself. What Happened in the Final Engagement on May 10, 1972? Cunningham and Driscoll had already downed two MiG 17s on the mission, bringing their career total to four, when a third MiG appeared in a head-on pass as Showtime 100 turned toward the coast. The resulting engagement lasted several minutes. The North Vietnamese pilot chose a vertical seesaw fight, an unusual tactic for the MiG 17, which was most effective in horizontal turning combat. Cunningham recognized the maneuver and committed to the vertical, forcing both aircraft through a series of zoom climbs and pitch-overs that progressively eroded the MiG’s energy state. At the apex of a zoom climb, both aircraft momentarily near stall speed with their noses pointed skyward, Cunningham cut throttle and deployed speed brakes. His deceleration was sharp. The MiG, with less drag and committed to the same upward vector, could not decelerate quickly enough. It overshot. In the language of Energy Maneuverability theory, Cunningham had reversed the engagement from defensive to offensive in a fraction of a second by manipulating his own energy state. He rolled in behind the MiG and fired his last AIM 9 Sidewinder. The missile struck. The aircraft disintegrated. On egress toward the Gulf of Tonkin, Showtime 100 was struck by a surface-to-air missile. Cunningham and Driscoll maintained control long enough to cross the coastline before ejecting over water. A United States Air Force search-and-rescue helicopter recovered them. Both received the Navy Cross. What Did the Navy Fighter Weapons School Actually Teach? The Navy Fighter Weapons School, established at Naval Air Station Miramar on March 3, 1969, under Commander Dan Pedersen, built its curriculum around Energy Maneuverability theory developed by Air Force Colonel John Boyd. Boyd had demonstrated mathematically that combat performance could be expressed as a function of specific excess power: the capacity of an aircraft to change its energy state faster than an opponent. This framework provided a rigorous basis for training decisions that had previously been made on intuition and experience. The core tactical insight TOPGUN delivered was this: the F 4 Phantom’s disadvantage in horizontal turning combat against the MiG 17 was not determinative. The F 4’s superior thrust-to-weight ratio could be exploited in the vertical plane, where a zoom climb would drain energy from the lighter MiG faster than from the F 4. A pilot who understood how to force the fight vertical, manage his own energy state, and use deceleration to induce an overshoot could defeat a platform that was categorically superior in horizontal maneuvering. That understanding required disciplined training against realistic opponents, not abstract theory. What Did the May 10 Mission Prove About Institutional Training? By 1972, with TOPGUN graduates distributed across fleet squadrons and teaching air combat maneuvering throughout the naval aviation community, the Navy’s kill ratio in Vietnam combat had risen to 13 to 1. No other variable introduced during the intervening three years accounts for a ratio change of that magnitude. Cunningham’s engagement on May 10 was the most visible demonstration of what the school’s curriculum produced under live conditions. He applied the specific tactics TOPGUN had developed and then returned to the school as an instructor, putting his combat experience directly back into the training pipeline before completing twenty years of naval service. The questions the May 10 engagement leaves open are worth tracking. Whether the kill ratio improvement is attributable primarily to TOPGUN instruction, to the changed rules of engagement under Linebacker, or to some combination of those and other factors has not been isolated in peer-reviewed scholarship. The identity of Cunningham’s final opponent that day remains permanently uncertain from available records. And the degree to which the Energy Maneuverability framework Boyd developed, which drove both TOPGUN’s curriculum and the aircraft design requirements that eventually produced the F 16 and F 18, continues to shape modern fighter pilot training is a question the next analysis in this series will address directly. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theinnovationattorney.substack.com/subscribe [https://theinnovationattorney.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

DARPA’s OTA Lets Startups Keep Their IP

DARPA’s Other Transaction Authority under 10 U.S.C. Section 4022 lets defense technology startups negotiate IP ownership with the government and avoid the default government license the Federal Acquisition Regulation imposes. An Other Transaction Agreement for a prototype project sits entirely outside the Federal Acquisition Regulation and outside the statutory data rights scheme that automatically attaches Unlimited Rights, Government Purpose Rights, or Limited Rights to a standard defense contract. Because none of those default categories apply unless the parties write them into the agreement, a startup negotiates the intellectual property schedule from a blank page rather than from a government favorable template. The startups that benefit most treat that blank page as a drafting opportunity rather than paperwork to sign quickly so a program can start. I spent three years as a submarine officer aboard USS William H. Bates before spending three decades advising technology companies, and one pattern has not changed in that time: the government’s leverage over intellectual property comes from a license structure most founders never read until a term sheet is already signed. Other Transaction Authority exists because Congress recognized, and expanded the authority repeatedly since, that a young company building autonomous flight software or a novel sensor has no reason to accept the same data rights scheme built for a shipyard constructing a destroyer under a fixed price contract spanning a decade. What does 10 U.S.C. Section 4022 actually let the Pentagon do? 10 U.S.C. Section 4022 gives the Pentagon authority to enter into transactions other than contracts, grants, or cooperative agreements for prototype projects directly relevant to enhancing the mission effectiveness of military personnel or the equipment the armed forces use. The statute sits beside 10 U.S.C. Section 4021, which covers research transactions, but Section 4022 is the provision startups encounter most often because it governs the prototype stage, where a working demonstration, not a research paper, is the deliverable. To use the prototype authority at all, at least one of several conditions must be met: a nontraditional defense contractor or nonprofit research institution must participate to a significant extent, every significant participant must be a small business or nontraditional contractor, at least one third of the cost must come from non federal sources, or the agreement must have used competitive procedures with at least two competing proposals. Agreements between one hundred million and five hundred million dollars require a written determination from the head of the contracting activity or the relevant DARPA, Defense Innovation Unit, or Missile Defense Agency director, and anything above five hundred million dollars needs approval from the Under Secretary of Defense for Acquisition and Sustainment along with notice to Congress. How does an Other Transaction Agreement change who owns the resulting technology? An Other Transaction Agreement changes ownership outcomes because it removes the automatic license structure a Federal Acquisition Regulation contract imposes and replaces it with whatever the parties negotiate. Under a standard defense award, the government receives Unlimited Rights in technical data or software developed solely at government expense, Limited Rights in data developed solely at private expense that embodies a trade secret, and Government Purpose Rights, which convert to unlimited after a five year period, in anything developed with mixed funding. Because 10 U.S.C. Sections 3771 through 3775, the statutes that create this scheme, do not apply to a prototype Other Transaction Agreement unless the parties import them by reference, a startup can instead retain its background intellectual property outright, grant a license limited to the specific data delivered rather than everything created during performance, or shorten the government purpose period well below five years. The founders who capture this advantage negotiate the intellectual property schedule as a standalone exhibit rather than accepting whatever boilerplate a contracting officer attaches to the solicitation, since nothing in the statute requires favorable terms to be offered first. What happens when a prototype moves toward full production? 10 U.S.C. Section 4022 allows a follow on production contract to be awarded without further competition, but only if three conditions are met: competitive procedures were used to award the original prototype agreement, the prototype project was completed successfully, and the follow on production option was written into the prototype agreement from the start. The Government Accountability Office tested exactly this sequence in Oracle America, decided in 2018, sustaining a protest because the Army’s prototype agreement contained no follow on production provision and the awardee had not finished the prototype before the follow on award was made. The lesson for a startup negotiating its first Other Transaction Agreement is to insist that the follow on production path, along with the intellectual property terms that will govern it, is written into the original agreement rather than left for later, since a company that waits until the prototype succeeds to negotiate production terms has already given up most of its leverage. Can a competitor stop the government from using this authority? A competitor’s ability to challenge an Other Transaction Agreement is narrower than it would be for a Federal Acquisition Regulation contract, though it is not eliminated entirely. The Government Accountability Office does not review the substance of who should have won an award, but it will review whether the statutory conditions for using the authority were satisfied, which is exactly the theory that succeeded in Oracle America. Commentary on a 2025 Court of Federal Claims decision suggests that court views itself as an available forum for these disputes too, so a startup cannot know with certainty today which forum will hear a future challenge. A startup that documents compliance with the statutory conditions at the time of award, rather than assuming the Other Transaction label makes the agreement immune from challenge, protects both its funding timeline and its negotiated intellectual property terms. Two questions remain live for any startup building a contracting strategy around this authority. A startup cannot know today which forum, the Government Accountability Office or the Court of Federal Claims, will hear a future challenge to its award, since the Court of Federal Claims has only recently begun describing itself as available for these disputes. A startup also cannot assume the intellectual property template it receives this year already reflects the broader nontraditional contractor exemptions Congress adopted in the fiscal year 2026 National Defense Authorization Act, since the Pentagon has not yet published updated model schedules. The next piece in this series looks at how consortium managers who administer Other Transaction Agreements on behalf of DARPA allocate intellectual property terms among member companies, and why a startup joining a consortium for the first time should read the consortium agreement before it reads the individual project agreement. Interested in analysis about the intersection of tech, policy and the law? Check out my Substack channel. https://theinnovationattorney.substack.com/ This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theinnovationattorney.substack.com/subscribe [https://theinnovationattorney.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]