OpenAI Shows Some Receipts

Well, we've established one thing: there are perhaps too many people with the name 'Chang' that work at both OpenAI and Apple.

I mean, that's really the connecting thread in the messages that OpenAI has now published publicly to defend themselves against Apple's lawsuit (and hoping to halt a would-be discovery process, which is clearly Apple's bigger angle here).¹ In both the iMessages posted between (former Apple employee and current OpenAI employee) Chang Liu and an unnamed current (at least at the time) Apple employee and in the emails posted between Apple outside counsel, Gabriel Gross and OpenAI’s General Counsel Che Chang, naming mix-ups play a central role.

We already knew this about the emails thanks to reporting by NBC News a few weeks ago. This just seemingly confirms that what happened is that Apple's outside counsel Gross, got mixed up when corresponding with an OpenAI employee named Wang and accidentally emailed an OpenAI employee named Chang in a follow-up response.

It just so happens that the Chang here is actually Che Chang, who just so happens to be OpenAI's General Counsel. So yeah, that's never a great mistake to make, especially for a lawyer sending legal documents, but extra especially when the mixup involves the GC of the other company.

But actually, reading between the lines of these messages, it seems Gross did intend to email Che Chang at least in that first correspondence at 5:53pm. The redactions obfuscate this, but Gross suggests that the 5:53pm email was intended for Chang, but the follow-up one at 6:07pm (actually 6:06pm, but I digress) was meant for Wang.

Given the emails, it reads like Gross emailed Chang to discuss the situation about Wang (and included the letter sent to him alongside other documents seemingly related to Apple information being mishandled). He then followed-up with Chang 13 minutes later, but meant to message Wang, because the two had seemingly just had a phone call. The fact that he still addresses him as "Mr. Chang" (when he should be addressing "Mr. Wang") just makes this all the more confusing (likely because he was replying to the wrong thread – how amazing would it be if AI auto-inserted that greeting?). But that's seemingly what happened.

But again, the fact that Chang is not some other random OpenAI employee but instead OpenAI's GC clearly sent this mistake in a new and interesting direction.

OpenAI's lawyer was clearly like "what the fuck is this?" and quickly emails Apple's actual in-house legal team to see if it's legit or some sort of weird phishing expedition. And if it is legit, he immediately seeks to get Gross removed from the account "given this behavior". Gross emails Chang the following day to explain the mix-up, and then Apple's in-house legal team does the day after that to confirm Gross is legitimate and noting that he should have already sent an email explaining the error. And that's that. At least, that's the last email OpenAI shared here. And from their preamble, that's the last correspondence.

OpenAI notes (twice) that they believed that Apple was "resolving any issues", but that was just the wording from Gross – again, the outside counsel – to Chang, so it seems a bit misleading to frame that as Apple closing out the matter entirely. But at the same time, the person named "Wang" is not a part of Apple's actual lawsuit, filed months later, so it is weird for Apple to use this back-and-forth as the framing for that lawsuit.

Also adding to the confusion here is that another Chang – Chang Liu – is one of the two key figures in the lawsuit filed by Apple, but seemingly has nothing to do with the email back and forth here. Instead, it's this other former Apple/current OpenAI employee named "Wang" who seemingly was being monitored for the same or similar Apple information handling/mishandling issues that Chang Liu is being accused of.

Got it? Because I barely do.

The fact that Apple framed this whole mix-up as OpenAI not responding to their initial inquiry is... not great. It does look like they were responding – well, at least Wang – and that at least that issue was being resolved. Apple is seemingly framing Che Chang not responding to actually talk with Gabriel Gross (in that first email) as the non-response bit, but again, that was sort of Apple's own fault! Well, their outside counsel's at least.

At the very least, it gave OpenAI an opening to not respond to the actual issue. But they probably still should have?! As OpenAI writes today:

Apple had claimed that they contacted OpenAI in February and that we didn’t respond. They now admit that their outside lawyers emailed the wrong person after confusing two Asian last names—only after we brought this to their attention. Apple also claimed they had a discussion with our General Counsel, which they now concede never happened. But they again hide the fact that they never raised the specific allegations in this lawsuit at that time, and that they in fact told us that they were “resolving any issues”. We then heard nothing for five months until they sued. In their latest filing, Apple tries hard to spin this sequence of events, but you can just read the emails for yourself here⁠.

This reads like a fairly disingenuous retelling of events given the emails OpenAI shared. And the company has some history of being slippery in this regard.²

At the same time, the next paragraph from OpenAI reads like a decidedly more fair and level-headed assessment of the situation (given the messages):

Apple accuses Chang Liu of accessing Apple confidential information after leaving the company, but only now admits that Apple employees reached out to him and asked for his help to locate this information (you can read the messages here⁠). Apple now tries to shift the blame to “residual access”, but they also don’t disclose that this is a common issue with Apple which is caused by them failing to properly manage system access when people leave. What that means in practice is that former employees who are trying to do the right thing when they leave still have access to Apple files—despite not wanting them or even being aware of them.

It is interesting/curious that this "common issue" was itself recently in the press as an "exclusive". Feels a bit strategically leaky...

Back to the iMessages,³ the current Apple employee clearly gets spooked when they get a notification that "Chang" has joined a spreadsheet. They note that they may have to "hit the workday button now" which suggests it would finally and fully cut off access to Chang Liu's files – which is fairly damning as it's an acknowledgement that they know they shouldn't be accessing these files and that they know how to turn it off. (And actually, it's Chang Liu who pushes to turn iMessage access off because it might leak OpenAI information the other way, to Apple!)

Chang Liu responds that he doesn't have access to the spreadsheet in question and that it's "Maybe the new Chang". Yes, it's seemingly another Chang naming mix-up!⁴ So they keep going...

Anyway, most of the iMessages that OpenAI shares here are decidedly more murky, simply because there's so much that's redacted. Obviously this is to protect Apple's confidential information – largely, it seems project names but also clearly some details on timing; it's impossible to know what else given the redactions – and it's still not a great look that an OpenAI employee is talking about such things so openly with a current Apple employee.

The most damning portion may be when other current Apple employees are added to the chat thread and one promptly responds with:

Hi, this is highly irregular, please remove me from this thread

Yeah, not great.

Then again, these messages do help frame it as much more of a casual correspondence with the Apple employee seeking help from their former colleague. Anyone who has ever left a job can undoubtedly relate. Someone you used to work with pings you for some help on something you used to know about/work on. Technically, you probably shouldn't help them, but come on, we're all humans. So yeah, these messages make it seem less like some big coordinated effort to steal from Apple and more like some bad judgement on a former Apple employee's part. But also the current Apple employee's part too!

In that regard, it's probably a good thing for OpenAI that this is angling towards a jury trial, because they might be more um, open to seeing the human error side of this. Because technically, legally, there's still clearly some at the very least proprietary information changing hands here.

And that's not great for OpenAI even if it just opens the door to discovery. More and more that seems to be the real angle here. Apple appears to be using this situation as a wedge to pry open the doors to any and all communication within OpenAI about information that relates to what former Apple employees are now working on within OpenAI.

Like, say, an AI speaker. Or a phone-like device.

You have to believe there's going to be a lot of such information simply given the fact that some 400+ people who used to work at Apple now work at OpenAI! That doesn't mean any of it will be damning let alone illegal, but well, Apple would like to see that to decide for themselves!

OpenAI would not like Apple to see any of that, one imagines.⁵

So the question now is if a judge is going to allow that discovery process to happen. It's clearly not a coincidence that OpenAI is releasing their side of the story now as yesterday Apple formally filed their request for a preliminary injunction against Chang Liu and Tang Tan (who OpenAI defended in their post today, but didn't release any documents/correspondence about). Most notably, Apple is also seeking "expedited discovery". And also seeking depositions of Chang Liu and Tang Tan and Yu-Ting Peng.

If you recognize that last name, it's because she's named in the suit as a co-conspirator with Chang Liu, but interestingly, wasn't actually sued by Apple. Given everything that has been reported, you'd be forgiven to think the messages OpenAI published here were between Chang Liu and Peng. But wait, at one point he says "Thank you sir!" in response which sure suggests it's not Peng. And actually, their potentially problematic correspondence was stated to have taken place over Line, not iMessage so... Why is OpenAI publishing this set of messages? Pretty clearly to establish that Chang Liu was simply trying to help Apple employees, who were the ones asking for it. But this clearly isn't the problematic communication and actions that Apple pointed to in the actual lawsuit – which also includes the downloading of documents. The fact that OpenAI didn't publish those now looks weird, at best.

"Just iMessages might get you in trouble," as Chang Liu writes at one point. For now, indeed. But there's clearly more to come. A lot more.

One more thing: per the Reuters report on Apple's preliminary injunction, beyond Tang Tan, Chang Liu, and Yu-Ting Peng, Apple also is specifically looking to depose another "unnamed OpenAI employee who previously worked at Apple." Unclear who that might be. Could it be "Wang"?! Perhaps even more intriguingly:

Apple also sought depositions from corporate representatives of OpenAI and ‌io ⁠Products, which is OpenAI's commercial arm and is listed as a defendant in the lawsuit.

Is this Apple's way to subtly rope in Jony Ive as well? Or are they going to continue to (clearly intentionally) leave him out of this? What about Sam Altman?

¹ Pretty standard OpenAI playbook at this point to publish such corresepondence, see also: the Elon Musk lawsuit. OpenAI always seems to have to "sadly" publish such things. Which in this case, did not stop a trial (and the embarassing disclosures) even if OpenAI won that one in the end.

² As an aside, is it too 3D Chess to think that Che Chang made such a big deal out of this mix-up knowing that it could help any potential legal issues down the road – and/or distract from what Apple's outside counsel is looking into? But then again, we don't know what documents were actually inadvertently shared, so... I'm just putting it out there! ↩

³ Which, kudos to OpenAI for the formatting in their blogpost. Not only does the back-and-forth actually resemble actual iMessages, there are blue bubbles! ↩

⁴ "That new Chang has big shoes to fill." ↩

⁵ Another reason why, perhaps, they shouldn't have poked the bear in the months leading up to this situation... ↩