Imagine building a company from scratch, pouring every waking hour into a product you believe in, only to find yourself in a federal courtroom facing a lawsuit that could shut it all down. That was the reality for Freightmate AI, a Bellevue, Washington-based logistics software startup, when San Francisco freight forwarder Flexport accused it of stealing trade secrets. The case has now been resolved, and for the small team at Freightmate, the relief is palpable. The two companies notified the court of a settlement on October 5, just two weeks before a jury trial was scheduled to begin in U.S. District Court in San Francisco before Judge Rita Lin. Under the terms of the deal, Freightmate’s products will remain on the market. Flexport had sought a court order that could have temporarily barred the startup from selling its main product, which would have been a devastating blow for a company with only 17 employees. Freightmate’s CEO and co-founder, Bryan Lacaillade, put it simply: “We’re glad to have this resolved.” In a statement, he thanked his team, customers, and partners for their support, and expressed excitement about what lies ahead. The financial terms of the settlement were not disclosed, but the stipulated judgment says Flexport’s claims for monetary damages were resolved in a confidential agreement, and each side will cover its own legal fees. For a startup that had raised only about $5.65 million in total funding, avoiding a lengthy trial and a potential injunction is not just a legal victory—it’s a chance to keep breathing.
The story behind the lawsuit begins with ambition and a familiar pattern in the tech world: talented employees leave a big company to start something of their own. Lacaillade and fellow co-founder Jason Zhao both joined Flexport from Amazon in 2021, and they left in 2024 to launch Freightmate, which came out of stealth that June with a $650,000 pre-seed round from Wischoff Ventures. The company later raised a $5 million seed round led by Fuse in January 2025. Freightmate sells AI-powered freight forwarding software, along with add-on tools that can pull data from shipping documents, handle customs paperwork, and process invoices. Its customers include some of the world’s largest freight forwarding and logistics companies. For a small startup, that kind of client list is a sign of real traction. The founders were not outsiders dabbling in an industry they didn’t understand; they were insiders who had spent years learning how global shipping works. That background made them credible to investors and customers, but it also made them vulnerable to legal scrutiny from their former employer. Flexport sued in March 2025, alleging that the founders secretly planned the startup while still working at the company. Lacaillade and Zhao were not alone in the venture—Rishab Gadroo was also a co-founder—and together they represented a classic Silicon Valley success story in the making, until the past caught up with them.
At the heart of the case was a question that is becoming increasingly common in the age of artificial intelligence: do everyday business documents count as valuable trade secrets when they can be fed into AI models? The dispute centered on shipping documents. Flexport alleged that Freightmate acquired approximately 2,000 confidential Flexport shipping documents and used some portion of them. These included bills of lading, arrival notices, and other paperwork from about 70 Flexport shipments. Court filings showed that Lacaillade and Gadroo each ran five to ten of the documents through OpenAI’s GPT-4o model in 2024 to test how well it could pull out standard information. That might sound like a small, almost technical experiment, but Flexport argued it was far more significant. The company claimed the documents helped shape the prompts Freightmate wrote for its AI system, giving the startup a head start in building its product. In other words, Flexport saw the use of those documents not as a minor misstep, but as the foundation of a competing product built on stolen knowledge. Freightmate responded that it built its product independently and, more boldly, that the documents were not trade secrets in the first place. This argument struck at an emerging vulnerability in the AI industry: if ordinary business paperwork can be classified as a trade secret simply because it was used to train a model, then every employee who leaves a company with documents in their inbox could become a target.
The settlement did not deliver a clear verdict on that question, but it did include some important admissions. As part of the stipulated judgment, Lacaillade and Zhao acknowledged that they and the company acquired approximately 2,000 confidential Flexport shipping documents and used some portion of them. Zhao specifically acknowledged that he acquired Flexport confidential information, including trade secrets, and both founders agreed that they knew or should have known that acquiring the shipping documents violated their agreements with Flexport. Those are significant words, and Flexport was quick to highlight them. Courtney Heger, Flexport’s vice president and head of marketing and communications, said the judgment includes key admissions that were backed up by a detailed forensic review conducted during the case, though the review’s findings were not made public. But the judgment stopped short of a finding of liability. Freightmate denied Flexport’s claim that its product was built on the documents, and the court ultimately made no finding on that point. Lacaillade was careful to emphasize that the stipulated facts were agreed to only as part of resolving the case, and only for the purposes of the lawsuit. He underscored that Freightmate and its founders deny liability, that nothing else in the judgment is an admission, and that the settlement preserves Freightmate’s right to compete and to independently develop and sell its software. It is a familiar legal dance: admitting certain facts to make a lawsuit go away while denying what those facts might imply. For the founders, it was likely a difficult choice between fighting a long, expensive battle and accepting language that could be used against them in reputation.
The practical consequences of the settlement are significant, but they are not catastrophic for Freightmate. Going forward, the judgment calls for a neutral forensic examiner to search Freightmate’s systems and cloud storage for Flexport documents. If any are found, the examiner will oversee their deletion, along with any product changes needed to remove Flexport data. That review is due to be completed by January 30, 2027, giving the startup several months to clean house under supervision. Lacaillade said Freightmate is cooperating fully with the review. For Flexport, this was a way to ensure that its confidential information is removed from the startup’s systems without the risk and uncertainty of a jury trial. For Freightmate, it means the company can continue operating, selling its software, and serving its customers, albeit with a forensic examiner looking over its shoulder for a period of time. The fact that Flexport did not secure an injunction is notable. It suggests that the court was not convinced that Freightmate’s entire product was tainted, or that Flexport’s harm could only be remedied by shutting the startup down. In a way, the settlement is a pragmatic compromise: Flexport gets acknowledgment and cleanup, and Freightmate gets to survive. Neither side got everything it wanted, but both can walk away and focus on their respective businesses. Flexport can return to its core freight forwarding operations, and Freightmate can continue building the AI tools that had attracted a promising roster of customers.
The broader significance of this case may extend far beyond the two companies involved. It highlights a question that will only become more pressing as artificial intelligence becomes embedded in every industry: when employees leave a company and use their knowledge, experience, and even documents to build AI-powered products, where is the line between legitimate competition and theft of trade secrets? Flexport’s argument was essentially that shipping documents, which might seem mundane to outsiders, are valuable because they contain structured data and patterns that can be used to train AI systems. That argument has implications for every company that handles sensitive or proprietary information. If feeding documents into an AI model is enough to make those documents trade secrets, then the barrier for proving trade secret misappropriation might be lower than many startups expect. On the other hand, Freightmate’s defense—that the documents were not trade secrets and that the product was built independently—reflects the reality that in a fast-moving AI landscape, many products are built using large language models trained on vast amounts of data. The line between using public knowledge and using confidential information is often blurry. For now, the settlement leaves that question unresolved. No court ruled on whether the shipping documents were trade secrets, and no court decided whether Freightmate’s product was actually built on them. What remains is a cautionary tale for founders who leave big companies with files in their possession, and a reminder that in the age of AI, even ordinary paperwork can become the center of a legal firestorm. Lacaillade said the startup’s focus is on its customers and on building the company. That may be the most human response of all: after months of legal turmoil, the best way forward is to simply get back to work.
