Apple sued OpenAI and it looks pretty bad

Hey Reader,

When I saw the headlines last week about Apple suing OpenAI, I figured it was another boring corporate controversy.

But after reading the details of Apple's allegations, it seems like a big deal – perhaps even more brazen than OpenAI switching from a charity to a for-profit, which spurred another big recent lawsuit.

It centers around OpenAI wanting to create hardware – such as phones, watches or wearable devices – that are specifically designed to run AI. They made a splashy announcement about this a year ago, which included them recruiting one of Apple's most famous hardware designers.

Right now, no hardware has been announced. But according to Apple, OpenAI has been systematically poaching ~400 Apple employees for its new hardware department and asking them to bring trade secrets with them.

Some examples from this Bloomberg report:

According to a lawsuit filed Friday, Liu departed with three things: a company-issued MacBook he never returned, a close relationship with an Apple employee who continued sharing internal information, and, most significantly, knowledge of a software bug that gave him ongoing access to internal file servers.

“LOL, I found out I can access the [network storage], so funny,” Liu wrote to his former Apple colleague, Alyssa Peng. He then, Apple alleges, used that access to download presentations, hardware designs, manufacturing details and testing procedures — all while already working at OpenAI.

When Liu discovered the bug, Peng stepped in to help, according to the suit. She replied, “I’m ready,” and eventually helped obtain more information through her own laptop. A few months later, in April, Peng herself left for OpenAI’s growing hardware division.

She joined more than 400 other former Apple employees drawn by the opportunity to work on next-generation devices designed to replace the two-decade-old iPhone — with salaries and stock options that trumped Apple’s less dazzling pay packages.

The episode is one of many that Apple says illustrate a “systematic effort to acquire, retain and use” its confidential information to help OpenAI replicate its decades of work building the world’s most successful consumer electronics business.

Yikes!

What I find fascinating about this is that it does not seem like Apple is relying on conjecture or circumstantial evidence – they seem to have real conversations in their own logs where former employees accessed Apple's confidential files while working at OpenAI.

Publicly, OpenAI is claiming the suit isn't a big deal – but my early impression is that it puts all their future hardware products in question, since pretty much anything they do could ostensibly be threatened by Apple's intellectual property claims now.

We'll have to wait months or years to see how this particular suit plays out. But, assuming Apple's claims are true, it seems like it's one more example of a pattern of the leadership of OpenAI acting without ethics, responsibility or honor.

Anthropic is a company full of people who quit OpenAI because they didn't like the brazenness of its leadership. OpenAI's board fired their own CEO (Sam Altman) because they believed he was untrustworthy, only to reverse the decision a few days later. Even Elon Musk, no paragon of honor himself, had a reasonable claim that OpenAI was wrong to switch from a non-profit to a for-profit (Musk ultimately lost his lawsuit on that point).

Ultimately, though, OpenAI may not really matter a few years from now.

That's the irony of ChatGPT being a household name. Not only is OpenAI probably the least trustworthy and least responsible AI company in general, they're also falling behind in almost every way (more on that in a moment). The result: despite their brand recognition, the AI industry might pass them by – sort of like what happened to TiVo in the age of DVRs and early streaming.

Here's the thing: you have more choices than ever when picking a large language model – and, at least at the moment, LLMs are all that OpenAI has to offer. (They're shutting down some side projects in advance of their IPO and haven't released any hardware yet.)

You could use an established company's LLM, like the excellent Google Gemini 3.1.
You could use the Claude line from Anthropic. The company is young but is currently the best in the business in terms of frontier models.
You could use an open-source, extremely inexpensive model like DeepSeek or Qwen, hosted on secure U.S.-based infrastructure. This is a little harder to set up, but many companies now give you the ability to use these super-cheap and super-good models in a way that is safe and doesn't touch any datacenters in China. To give you a sense of how cheap this is, Claude's Haiku model costs $5 per million tokens and the comparable DeepSeek v4 Flash costs 18¢ per 1 million tokens. That's a 27x price difference for a very similar outcome with no security risk.

Meanwhile, pretty much everyone else (Meta, Amazon, etc.) is trying to build even better, cheaper models.

So, if no one trusts OpenAI, and they keep doing things that reinforce the idea that they are untrustworthy, and they no longer have the best LLMs...

What exactly is their plan?

Needless to say, OpenAI deserves credit for the huge breakthrough they made with GPT-3.5 in 2022, and they've released a lot of great ChatGPT products since then. But will customers stick around?

It's hard for me to see how OpenAI thrives in a world in which LLMs are a commodity and honor and trustworthiness are scarce.

Until next time,

Rob Howard
Founder, Innovating with AI

•

Our new membership, Ethical AI Foundations, is the most accessible way to learn AI responsibly. Two monthly workshops designed for any budget or schedule – all built on our C.A.R.E promise:

Current
Actionable
Responsible
Empowering

Our next workshop is AI for Good: Helping Your Community & Causes. Start today with a 30-day money-back guarantee.

👉 Join Ethical AI Foundations