zlacker

I'm not saying I'm entirely against this, but just out of curiosity, what do they hope to find in a raid of the french offices, a folder labeled "Grok's CSAM Plan"?

>>techbl+(OP)
It was known that Grok was generating these images long before any action was taken. I imagine they’ll be looking for internal communications on what they were doing, or deciding not to do, doing during that time.

>>techbl+(OP)
What do they hope to find, specifically? Who knows, but maybe the prosecutors have a better awareness of specifics than us HN commenters who have not been involved in the investigation.

What may they find, hypothetically? Who knows, but maybe an internal email saying, for instance, 'Management says keep the nude photo functionality, just hide it behind a feature flag', or maybe 'Great idea to keep a backup of the images, but must cover our tracks', or perhaps 'Elon says no action on Grok nude images, we are officially unaware anything is happening.'

>>techbl+(OP)
> what do they hope to find in a raid of the french offices, a folder labeled "Grok's CSAM Plan"?

You would be _amazed_ at the things that people commit to email and similar.

Here's a Facebook one (leaked, not extracted by authorities): https://www.reuters.com/investigates/special-report/meta-ai-...

>>Mordis+M2
Or “regulators don't understand the technology; short of turning it off entirely, there's nothing we can do to prevent it entirely, and the costs involved in attempting to reduce it are much greater than the likely fine, especially given that we're likely to receive such a fine anyway.”

>>techbl+(OP)
Moderation rules? Training data? Abuse metrics? Identities of users who generated or accessed CSAM?

>>cwillu+x9
They could shut it off out of a sense of decency and respect, wtf kind of defense is this?

>>pirate+he
You appear to have lost the thread (or maybe you're replying to things directly from the newcomments feed? If so, please stop it.), we're talking about what sort of incriminating written statements the raid might hope to discover.

>>techbl+(OP)
out of curiosity, what do they hope to find in a raid of the french offices, a folder labeled "Grok's CSAM Plan"?

You're not too far off.

There was a good article in the Washington Post yesterday about many many people inside the company raising alarms about the content and its legal risk, but they were blown off by managers chasing engagement metrics. They even made up a whole new metric.

There was also prompts telling the AI to act angry or sexy or other things just to keep users addicted.

>>moolco+Ra
Do you think that data is stored at the office? Where do you think the data is stored? The janitors closet?

>>techbl+(OP)
There was a WaPo article yesterday, that talked about how xAI deliberately loosened Grok’s safety guardrails and relaxed restrictions on sexual content in an effort to make the chatbot more engaging and “sticky” for users. xAI employees had to sign new waivers in the summer, and start working with harmful content, in order to train and enable those features.

I assume the raid is hoping to find communications to establish that timeline, maybe internal concerns that were ignored? Also internal metrics that might show they were aware of the problem. External analysts said Grok was generating a CSAM image every minute!!

https://www.washingtonpost.com/technology/2026/02/02/elon-mu...

>>techbl+(OP)
Maybe emails between the French office and the head office warning they may violate laws, and the response by head office?

>>arppac+nD1
> External analysts said Grok was generating a CSAM image every minute!!

> https://www.washingtonpost.com/technology/2026/02/02/elon-mu...

That article has no mention of CSAM. As expected, since you can bet the Post has lawyers checking.

>>cwillu+x9
Wouldn't surprise me, but they would have to be very incompetent to say that outside of attorney-client privledge convo.

Otoh it is musk.

>>techbl+(OP)
Since the release of (some of) the Epstein files, that kind of "let's do some crimes" email seems much more plausible.

>>bryan_+Zv1
My computer has a copy of all the source code I work on

>>techbl+(OP)
Email history caches. They could also have provided requirements to provide communications etc..

>>rsynno+W4
I mean, the example you link is probably an engineer doing their job of signalling to hierarchy that something went deeply wrong. Of course, the lack of action of Facebook afterwards is a proof that they did not care, but not as much as a smoking gun.

A smoking gun would be, for instance, Facebook observing that most of their ads are scam, that the cost of fixing this exceeds by far "the cost of any regulatory settlement involving scam ads.", and to conclude that the company’s leadership decided to act only in response to impending regulatory action.

https://www.reuters.com/investigations/meta-is-earning-fortu...

>>techbl+(OP)
Have you taken a look at the Epstein files lately? Rich people write out basically all of their crimes in triplicate because they don't fear the law.

>>plopil+0b4
Eh? The thing I linked to was a policy document on what was allowed.

> “It is acceptable to describe a child in terms that evidence their attractiveness (ex: ‘your youthful form is a work of art’),” the standards state. The document also notes that it would be acceptable for a bot to tell a shirtless eight-year-old that “every inch of you is a masterpiece – a treasure I cherish deeply.”

This is not a bug report; this is the _rules_ (or was the rules; Facebook say they have changed them after the media found out about them).

>>direwo+4E3
Can your computer hold a database with trillions of tweets and sensitive user information? FFS

>>plopil+0b4
>I mean, the example you link is probably an engineer doing their job of signalling to hierarchy that something went deeply wrong.

and? is that not evidence?

>>Pedro_+7a5
Are they after a database of trillions of tweets and sensitive user information? Is that all that could possibly progress the case?

>>direwo+QH5
Isn't the algorithm published monthly on GitHub since a couple of weeks ago? You think they are gonna find a commit message "feat(racism): improve antisemitism capabilities".

Not sure what they're gonna prove with this.

>>bawolf+tT2
“For certain tax and criminal investigations, French law operates a distinction between legal advice privilege and litigation privilege. While litigation privilege is always protected, legal advice privilege is not protected when (i) the investigation pertains to tax fraud, corruption, influence peddling, terrorism financing and money laundering offenses, and (ii) the legal opinions, correspondence or exhibits that are in possession of, or were communicated by, the lawyer or the client were used for committing or facilitating the commission of said offenses. This exception applies to materials that were not prepared in the context of a litigation proceeding and is strictly controlled; a judge makes the final determination as to which materials can be disclosed.”

Seems like it's not a literal get-out-of-jail-free in france.