Recent Posts

Apple's trade secret suit against OpenAI and OpenAI's own IPO reckoning are the same story, told from two directions.

In 2010, Steve Jobs told his biographer he intended to wage a "thermonuclear war" on Google's Android, a system he believed had been built by copying the iPhone. The war took the shape of patent suits against Samsung and other handset makers, an eight-year campaign that finally settled in 2018. Android had already become the dominant mobile operating system in the world by then. Apple's litigation never reversed that outcome. What it bought was time, months and years during which a competitor had to answer subpoenas and depositions instead of building.

That same instinct resurfaced on July 10, when Apple filed a 41-page federal complaint against OpenAI in the Northern District of California. The suit names OpenAI, its Chief Hardware Officer Tang Tan, and former Apple engineer Chang Liu, and alleges a scheme operating, in the complaint's own words, "at every level" of the company. Fourteen months earlier, this would have read as improbable. ChatGPT had been woven into iOS since 2024, and the working assumption inside both companies was a kind of division of labor: Apple would keep building the device, OpenAI would keep supplying the intelligence running on top of it.

That assumption broke down for a specific, traceable reason. In 2025, OpenAI acquired io Products, the hardware startup founded by former Apple design chief Jony Ive, in a deal Apple's complaint values at roughly $6.5 billion. Ive's involvement gave OpenAI's hardware ambitions a credibility no competitor's device push had carried before. OpenAI then hired Tan, who had spent 24 years at Apple and risen to vice president of product design for the iPhone and Apple Watch, according to TechCrunch's reporting on the filing. Apple alleges that Tan asked job candidates still employed at Apple to bring actual hardware components into interviews for what the complaint calls "show and tell" sessions, and used Apple's internal project codenames to draw out further detail, a claim detailed in Fortune's coverage of the suit. A second defendant, Liu, is accused separately of keeping an Apple-issued laptop after leaving the company and using a previously unknown authentication flaw to reach Apple's internal network for months afterward.

OpenAI's response was brief. "We have no interest in other companies' trade secrets," the company said in a statement. "We remain focused on building innovative technology that empowers people everywhere." Sam Altman, writing separately on X, said he respected Apple and was not afraid of it. None of Apple's allegations have been tested in court or proven, and OpenAI disputes them. But the filing accomplishes something regardless of how it resolves. It puts a legal cloud over the exact program OpenAI is counting on to reach the people who have never opened ChatGPT: a physical device meant to succeed the phone the way the phone succeeded the desktop. It also lands as Apple has already been quietly walking away from the partnership on its own terms, having turned to Google's Gemini rather than OpenAI's models to power the redesigned Siri arriving this fall.

This is the second front of a race we tracked in our last dispatch. The day before Apple's filing, OpenAI's ChatGPT Work and Meta's Muse Spark 1.1 turned the conversation about agentic AI from a software story into a workplace one: systems that plan a task, execute it, and report back rather than simply answering a question. Apple's lawsuit is evidence that the same competitive pressure now has a physical dimension. An agent that can act inside your calendar is valuable. An agent embedded in a device carried everywhere is a different order of business, and it is the one Apple has spent fifteen years building a moat around.

The timing compounds a second story that was already unfolding the same week, one with less drama but arguably higher stakes. OpenAI has spent recent months preparing what could be the largest technology listing in history, having filed confidential IPO paperwork with U.S. regulators. It was not the first to file. Anthropic, the AI lab founded by former OpenAI researchers, submitted its own confidential registration on June 1, ahead of OpenAI's own confidential filing later the same month, after a funding round that valued the company at $965 billion, above OpenAI's own $852 billion mark from earlier in the year. Anthropic has also told investors it was on track for its first operating profit in the quarter ending in June, according to reporting that traces back to the Journal itself, though the company has cautioned the figure is flattered by a temporary compute discount and may not repeat.

None of this alone would derail a listing. Together with the Apple suit, it changes the calculation OpenAI's board is making. A company preparing a public prospectus is obligated to disclose material legal risk, and a complaint alleging that a core growth program rests on misappropriated trade secrets is precisely that category of risk. OpenAI is reshuffling its leadership and reworking its product lineup at the same time.

Read on its own, each of these is a manageable setback. Read together, they describe an organization whose growth has outpaced its capacity to explain itself, first to a court, and shortly to public market investors who will expect the same discipline any listed company owes its shareholders every quarter. Whether it is a multi-billion-dollar lawsuit over untracked employee actions and hardware files, or a mid-market algorithm operating without oversight, the root failure is identical: a lack of internal controls and documentation. In both cases, the organization cannot produce a defensible record of what its people and systems did, who touched sensitive material, and on whose authority. If the biggest AI pioneers in the world cannot document who accessed what and when, mid-market businesses deploying agentic AI face an even greater risk.

This is the discipline Occams.ai was built to bring to organizations at any scale, well before a courtroom or a prospectus forces the question. Transparent reasoning behind every automated decision. Human accountability engineered before a workflow goes live, not added after something goes wrong. Documentation built to survive an audit, a regulator, or a curious investor, because eventually one of them asks.

Before any of them asks you to prove how your AI workflows operate, let us help you map your systems. Connect with our team for an initial 30-minute AI Exposure Assessment to identify where your governance is strong and where your liabilities are hidden.