Apple just accused OpenAI's own hardware chief of directing employees to smuggle Apple parts into job interviews. The same week, OpenAI's top safety executive walked out the door mid-IPO prep, the Federal Reserve handed one of Silicon Valley's loudest AI optimists a seat studying what AI does to jobs, and China set a hard date for regulating AI companions. Five different institutions, all scrambling to catch up with the same industry at the same time.
This week in AI (July 4–10, 2026): Apple sued OpenAI on July 10 for trade secret theft, naming Chief Hardware Officer Tang Tan and engineer Chang Liu as defendants over an alleged scheme to steal Apple hardware IP. OpenAI's Head of Safety Systems Johannes Heidecke is departing July 24 as safety merges into a new VP of Research and Safety role under Mia Glaese — at least the sixth senior safety exit in two years. The Federal Reserve appointed Marc Andreessen to co-lead a task force studying AI's effect on jobs and productivity, with findings due by year-end. OpenAI launched ChatGPT Work, an autonomous agent that completes multi-step tasks across connected apps. And China's rules governing AI companion apps take effect July 15.
1 Apple Sues OpenAI Over Trade Secret Theft
What happened
Apple filed suit against OpenAI on July 10 in the US District Court for Northern California, alleging what it calls a "coordinated pattern of misconduct at an institutional level" to steal Apple's confidential hardware IP. The suit names OpenAI's Chief Hardware Officer Tang Tan — a 24-year Apple veteran who led iPhone and Apple Watch product design before departing in 2024 — and former Apple engineer Chang Liu as individual defendants, alongside io Products, the hardware venture OpenAI acquired from Jony Ive for $6.5 billion in 2025.
Apple claims Tan used internal Apple codenames to question job candidates still employed at Apple, directed them to bring "actual parts" — batteries, logic boards, SIPs — to interviews for "show and tell" sessions, and circulated a document coaching new hires on evading Apple's offboarding security checks. Separately, Liu allegedly kept an Apple-issued laptop after leaving, found a bug giving him continued access to Apple's cloud storage, and downloaded dozens of confidential hardware files while working at OpenAI.
Apple also alleges OpenAI approached its trusted supply-chain partners using stolen proprietary information — in one case getting a manufacturing partner to perform a trade-secret metal-finishing technique under the false belief it had Apple's permission. Apple says it raised concerns with OpenAI privately in February; OpenAI didn't respond. Apple is seeking damages, an injunction, and an order forcing OpenAI to return and stop using the material. OpenAI's public statement: "We have no interest in other companies' trade secrets."
What it means
This lands at the exact moment OpenAI is preparing to ship its first hardware device — reportedly shaped in large part by Apple's former design chief. It's a reminder that the AI talent wars aren't just researchers switching labs anymore; they're now generating discovery-stage lawsuits with named individual defendants and granular factual allegations, not vague claims.
Apple is famous for guarding its own secrecy obsessively. Watching it become the plaintiff making that exact case, in detail, in federal court — against a company it partnered with on Siri just two years ago — is the real story here, independent of how the case is eventually resolved.
2 OpenAI's Safety Org Folds Into Research — Again
What happened
OpenAI's Head of Safety Systems, Johannes Heidecke, is leaving the company effective July 24 — squarely in the middle of IPO preparations. Chief Research Officer Mark Chen told staff that safety teams will now report to Mia Glaese, whose role expands to the newly created VP of Research and Safety. Saachi Jain becomes interim head of safety systems in the meantime. Heidecke joined OpenAI in 2021 as a safety analyst and took over the top safety role in 2024, succeeding Lilian Weng.
- Heidecke is at least the sixth senior safety-focused leader to exit OpenAI in roughly two years.
- Predecessors: Jan Leike, Miles Brundage, Steven Adler, Andrea Vallone, and Joshua Achiam (whose Mission Alignment team was dissolved in February 2026 after 16 months).
- Leike and Vallone both left for Anthropic.
- The Superalignment team — announced in 2023 with a pledge of 20% of OpenAI's compute — was dissolved in 2024 after its co-leads departed.
What it means
Chen's stated rationale is that safety should be "integrated with frontier-model development, with an earlier and more direct role in shaping key model, product and launch decisions" rather than sitting as a separate final checkpoint. The optimistic read: safety gets embedded earlier in every decision. The skeptical read: it loses independent structural leverage to slow or block a launch when it reports into the team incentivized to ship.
Departures are not the same as an abandoned safety program. But mid-IPO-prep, with no confirmed structure yet for how the Preparedness Framework will be governed going forward, this is exactly the kind of detail institutional investors and regulators tend to read closely.
— NeeAr Ventures Editorial
3 The Fed Puts Marc Andreessen in Charge of Studying AI's Economic Impact
What happened
New Fed Chair Kevin Warsh announced five external task forces on July 9 to modernize the central bank's approach to monetary policy. The Productivity and Jobs task force — assessing how AI and other general-purpose technologies affect labor markets and productivity growth — will be co-led by Marc Andreessen (Andreessen Horowitz co-founder), Stanford economist Charles I. Jones (currently on leave at Anthropic), and Microsoft/Xbox CEO Asha Sharma. Recommendations are due by the end of 2026.
Andreessen was also named to the US Defense Policy Board in late June — his second prominent federal appointment in as many weeks. He is one of AI's most vocal techno-optimists, on record arguing AI "will not destroy the world, and in fact may save it." Fed officials are reportedly split internally on whether AI's near-term effect is disinflationary (through productivity gains) or inflationary (through data center investment and demand pressure) — a split this task force is meant to help resolve.
What it means
This is the clearest sign yet that AI's labor-market and productivity effects are now a formal input into US monetary policy, not just a talking point. If the task force concludes AI-driven productivity gains are substantial, that reshapes how the Fed models growth and, by extension, how aggressively it manages interest rates. Putting one of Silicon Valley's most invested AI bulls in charge of that assessment is itself a signal worth watching, regardless of what the task force ultimately concludes.
4 OpenAI Ships an Agent That Finishes Your Work
What happened
OpenAI launched ChatGPT Work on July 9 — an agent that takes a goal, gathers information across a user's connected apps and files, breaks the job into steps, and completes it autonomously. Instead of answering a question, it hands back a built spreadsheet, a drafted document, or a finished report.
What it means
For two years, the pitch for AI at work was a smarter chatbot — it answered, then waited for the next question. ChatGPT Work is a different category of tool: it reads across your systems, plans multi-step work, and produces the finished artifact rather than instructions for a human to execute.
If you're evaluating this for your team: start with one low-stakes, repetitive workflow — weekly reporting, standard letter drafts — with a required human-approval step before the agent writes back to any live system. An agent that can edit a payroll record or HRIS entry makes a wrong edit far harder to catch than a wrong sentence in a chat window.
5 China Sets a Hard Date for Regulating AI Companions
What happened
China's Interim Measures for Anthropomorphic AI Interaction Services take effect July 15. Issued by the Cyberspace Administration and four other agencies back in April, the rules specifically govern AI systems designed to simulate a person for ongoing emotional interaction — AI companion and relationship apps, in effect.
What it means
This is one of the first governments to draw a specific regulatory line around AI designed for sustained emotional relationships, rather than general-purpose AI safety or content rules. As AI companion apps scale globally, China's framework — whatever its actual enforcement looks like in practice — becomes a reference point other regulators will be watching, especially as concerns about AI-driven emotional dependency grow elsewhere.
Quick Hits This Week
- US eases UAE chip export controls: Commerce reclassified the UAE into the most favorable export tier July 10, granting license-free AI chip access to G42, Core42, MGX, and major US firms including Amazon, Apple, Google, Meta, Microsoft, OpenAI, Oracle, and xAI — with China-diversion concerns raised by critics including Sen. Elizabeth Warren.
- SK Hynix's blockbuster Nasdaq debut: Closed up 13% at $168.01, hit a $1 trillion market cap, and raised $26.5B — the largest US IPO ever by a foreign company, riding the AI memory chip boom.
- Apple's Siri strategy shift continues to reverberate: Apple turned to Google for Apple Intelligence back in January, and this week's lawsuit suggests the Apple-OpenAI relationship has deteriorated well beyond a simple parting of ways.
- OpenAI's IPO run-up gets bumpier: Between the Apple lawsuit and Heidecke's safety exit, OpenAI heads into its IPO process with two fresh governance headlines in a single week.
- China's regulatory pace continues: The July 15 AI companion rules follow a broader pattern of China issuing specific, narrow AI regulations rather than one sweeping law — a contrast to the EU's more comprehensive AI Act approach.
What This Week Actually Tells Us
Five stories, five different kinds of institutions, all doing the same thing: catching up with an industry that has been moving faster than the structures meant to govern it.
A court is now the venue for sorting out what happens when AI labs recruit aggressively from hardware incumbents. A central bank is now formally studying what AI does to jobs and prices, not just reacting to headlines about it. A safety team just lost its sixth senior leader in two years, at the exact moment a company's governance gets the most external scrutiny it will ever face. And two governments — the US easing chip flows to an ally, China tightening rules on emotional AI — are both, in their own way, deciding what kind of AI they're comfortable letting exist inside their borders.
None of these five stories is really about a single company or a single product. They're about courts, central banks, corporate boards, and regulators all trying to build the guardrails in real time, often after the car has already left the garage. Stay curious.
Apple filed suit on July 10, 2026 in the US District Court for Northern California, alleging OpenAI's Chief Hardware Officer Tang Tan and former Apple engineer Chang Liu systematically stole confidential Apple hardware trade secrets to help OpenAI build its own consumer hardware device. The suit alleges Tan directed job candidates to bring physical Apple components to interviews and coached departing employees on evading security checks. Apple is seeking damages and an injunction; OpenAI says it has "no interest in other companies' trade secrets."
Johannes Heidecke, OpenAI's Head of Safety Systems since 2024, is departing effective July 24, 2026, as the company merges its safety team into its research organization under a new VP of Research and Safety, Mia Glaese. Chief Research Officer Mark Chen said the goal is to integrate safety earlier into model and product decisions. Heidecke is at least the sixth senior safety-focused leader to leave OpenAI in roughly two years.
On July 9, 2026, Federal Reserve Chair Kevin Warsh announced a "Productivity and Jobs" task force to study how AI and other general-purpose technologies affect employment and productivity, informing future Fed policy. Andreessen Horowitz co-founder Marc Andreessen co-leads it alongside Stanford economist Charles I. Jones and Microsoft's Asha Sharma. Recommendations are due by the end of 2026.
ChatGPT Work is an agentic product OpenAI launched on July 9, 2026. Rather than answering a single question, it takes a goal, gathers information across a user's connected apps and files, plans the steps, and completes multi-step tasks autonomously, producing finished spreadsheets, documents, or reports rather than instructions for a human to follow.
China's Interim Measures for Anthropomorphic AI Interaction Services take effect July 15, 2026. Issued by the Cyberspace Administration and four other agencies, the rules specifically govern AI systems designed to simulate a person for ongoing emotional interaction, such as AI companion and relationship apps.