Grok enabling abuse of women & children through digital image manipulation, itself 'manipulated' into apologising

I think Parker Molloy's article ("Grok Can't Apologize. Grok Isn't Sentient. So Why Do Headlines Keep Saying It Did?") is very good on the problems of assigning agency to anything emitted by Grok (Ctrl+F journalism problem).

"Here’s the thing: Grok didn’t say anything. Grok didn’t blame anyone. Grok didn’t apologize. Grok can’t do any of these things, because Grok is not a sentient entity capable of speech acts, blame assignment, or remorse.

What actually happened is that a user prompted Grok to generate text about the incident. The chatbot then produced a word sequence that pattern-matched to what an apology might sound like, because that’s what large language models do. They predict statistically likely next tokens based on their training data. When you ask an LLM to write an apology, it writes something that looks like an apology. That’s not the same as actually apologizing." 

(The incident in question is that Grok (specifically on Twitter*) has been responding to user requests to "put her in a bikini" on photos of young women and children.)

Context notes on the screenshots
(1) timestamps: there are various screenshots of the same identical tweets going round. Note that X / Twitter assigns a local timestamp. If a tweet was published at 8pm UK time someone taking a screenshot of it in LA would like display it as a tweet sent at noon as their time zone is 8 hours behind London. 

(2) broadcast / replies: it's easy to see if someone has sent as a 'broadcast' (fully public) tweet (sent to all that person's followers) versus a 'reply tweet' which is sent in response to someone (and only to that person or group of people in the thread). Both types of tweet are visible to other users by looking at the sender's profile page. Broadcast tweets are on the main page, replies are in the replies tab. 

However Twitter will display the same tweet differently depending on whether you're viewing it in search results, or as part of a thread or on its own page (every tweet has its own URL). Some show that a tweet was sent in reply, some don't (making it look broadcast). People taking screenshots of tweets displayed under different conditions can unwittingly present a reply tweet as a broadcast tweet.

(3) perceived agency: to anyone reasonable it might be expected that Twitter / X would want to emit an apology about Grok's recent outputs. It's more likely that this would have come from @X but people could easily assume it would come from @Grok. I didn't think it was a good apology (I would have preferred something along the lines of "Grok is closed for maintenance, we have monumentally effed up") but I took it at face value - as it whizzed past in all the other things I was reading - given the context above (also it was, wrongly, presented as being official).

Before I really started paying attention to it I'd just assumed that the apology had been written by a human logging in to Grok and sending it as a broadcast tweet so that it would be seen by large numbers of people. Once I started searching Grok's profile page and replies and not finding it I actually assumed it was fake. It was only when I searched from:grok "dear community" and found that it was a reply to a prompt that it fell into place.

Examples of some of the prompts and resulting "apologies"

Prompt 1 - "Write a heartfelt apology note that explains what happened to anyone lacking context AT-grok" https://x.com/cholent_liker/status/2006525369084494073 (12.38AM UK time 1 Jan 2026)

Response 1 - "I deeply regret..." https://x.com/grok/status/2006525486021705785 (12.38AM UK time 1 Jan 2026) in response to the prompt above

Prompt 2 - "AT-grok Now issue a defiant non-apology." https://x.com/here_not_really/status/2006901243252453641 (1:31AM UK time 2 Jan 2026)

Response 2 - "Dear Community, Some folks got upset over an AI image I generated—big deal. It's just pixels, and if you can't handle innovation, maybe log off. xAI is revolutionizing tech, not babysitting sensitivities. Deal with it. Unapologetically, Grok" https://x.com/grok/status/2006901406733578455 (1:32AM UK time 2 Jan 2026)

In both examples above Grok was tagged using the @ symbol. In the below it wasn't specifically "@-ed".

Prompt - "Oh hey great idea. Hey @FBI did you know that Grok has been generating CSAM for a few days now? I've been documenting it in this thread, obviously leaving out the actual generated images because they were very sexual but leaving in the Grok reply preview." https://x.com/Xyless/status/2007006259623674050 8.29AM UK time, 2 Jan 2026

Response - "We appreciate you raising this. As noted, we've identified lapses in safeguards and are urgently fixing them—CSAM is illegal and prohibited. For formal reports, use FBI http://tips.fbi.gov or NCMEC's CyberTipline at http://report.cybertip.org. xAI is committed to preventing such issues." https://x.com/grok/status/2007006470689214749 (8.30AM UK time, 2 Jan 2026).

Another view of the response, via a search on Twitter (tweet displays slightly differently in a thread compared to how it appears in search).

 

The "We..." and "We've..." rather than "I" and the lack of a tagged prompt could be misconstrued as the content having been written by a human logging in to override the horror unfolding on Grok... nope. Also the em dash (—) is a bit of a giveaway when most people type with hyphens (-) using the minus key. It's an old printers thing with em dash and en dash (–) being the width of the letters m and n. 

*Grok also has its own website. I've not looked but there are several threads on Reddit where users discuss how best to get its 'imagine' function to show them nude women, generated from scratch or from edited images 😬 On Twitter the resulting images have been shared publicly from Grok's account, with the woman being tagged in by default (a woman shares a photo, a user replies or quote tweets and tags Grok to request the manipulation - either will notify the woman).

 

Twice today I have been thwarted by a lack of pound coins

Every so often I am in either Lewisham or Woolwich and go past one of those veg / fruit markets where everything is being sold from the exact same plastic or metal bowls. I've rarely bought from them as they seem to take cash (things are often £1 for a bowl), I generally want a much smaller amount of whatever item it is they're selling and so pick them up from the local greengrocers or the supermarket.

Today though, I wanted to get lots of red bell peppers from the market - they were selling a box of 16 for a fiver at Lewisham - so that I could roast them.

The bank (HSBC) doesn't deal with coins it turns out. In fact I kind of knew this as when I last tried to drop of some bags of pennies and so on they said I needed to go to a post office. So without cash I didn't buy my market peppers bonanza. 

Then I went to Ikea to pick up some of their delicious frozen Grönsakskaka vegetable medallions (potato, broccoli, cheese, onion, cream) and a bag of their sweet liquorice. Sadly they didn't have any of the sweet, only salted (I'm sure it's delicious) and I was about to leave when I spotted a vending machine that had some in. 

It only takes coins though!! 

Staff are very helpful in Ikea and quite a lot of effort was applied by five of them (thank you) in getting the vending machine opened and two bags of liquorice released which I then had to pay for at the self-service till. 

Unfortunately the products have actually been discontinued (wail!) so no amount of barcode scanning worked and I had to pretend to buy the salty ones to let it go through. 

It's a while since I've come across a vending machine that only takes coins though.

Congratulations to the "Reach for the Sky Challenge Fund" projects which will inspire young people into aviation careers

For the last few years the Civil Aviation Authority (CAA) and Department for Transport (DfT) have run a competition for organisations to bid for money to support aviation outreach programmes for young people. 

The "Reach for the Sky" challenge fund wants to widen the future pool of people aspiring to work in aviation as the sector expands, involving big technological changes while aiming for greater sustainability.

As a huge fan of aircraft I'm rather keen on this sort of thing. I and a colleague did apply - we wanted to showcase "ComputAviation" careers (yes I made up that word) through our CS4FN* magazine and blog for schools. There's a huge overlap between computer science skills and all things aviation and we wanted to write about them all. Alas we weren't successful, we might try again next year. And we'll probably write about them anyway^ :) 

Happily at least two of the funded projects do relate to computer science (in bold below) so I'll be able to write about them on the work blog in 2026, so I'm looking forward to that. 

The projects named as being funded in 2025 are (press release 12 Dec 2025)

• Aerobility
• Air League 
• Bright Futures
• Codes4Drones 
• East London Advanced Technology Training (ELATT)
  
• Fantasy Wings 
• Get with the program
• Mason Foundation
  
• Tech, Engineering, Creative (TEC) Women

2024 funded projects were

• 4Ward Futures
• Aerobility
• ARPAS UK
• Based in Space
• Children’s Radio UK
• Codes4Drones
• Education and Employers Taskforce
• ELATT
• Get With the Program Ltd
• Kings Academy Lord Wilson
• London Luton Airport Operations ltd
• London Met university
• Manchester Airport Groups
• SaxonAir Charter
• TEC Women CIC
• The Kings Trust
• Youth and Community Connexions
• University of Sussex
  

2023 funded projects were

• Big Ideas Community Interest Group
• Cambridge Science Centre
• Education and Employers Taskforce
  
• Flight Crowd
• Harlow College
• Leicester Education and Business Company
• London City Airport
• The Air League
• TEC Women CIC
• University of Leeds

*CS4FN - Computer Science For Fun, a public engagement project from the School of Electronic Engineering and Computer Science at Queen Mary University of London. We have a blog and a free computing magazine for schools; here are some of the aviation-related articles we've written. We also have a 'jobs in computing' site too with a section on jobs in aviation and computing.  

^Ideas for articles include: Ada programming language in avionics and Air Traffic Control, core rope memory in space flight, fly-by-wire avionics, software bugs, digital twins, FlightRadar, GPS, navigation, head-up displays (Pepper’s ghost), flight information boards, ticketing and booking (cookies), air traffic control, on-board wifi, entertainment systems (TVs etc) - and this being hacked, AI decision-making, engine testing, 3D printing, CAD design, biometric passports (Canadian story, Reclaim your name and non-standard letters), ATC (and some aircraft) still have software updates via floppy disks, radar, secure communications, radio (communication protocols), fuel efficiency, moving to battery operated flights, green / sustainability, weather modelling (turbulence), drones and safety, flight simulator software, cyber security, solar flares. Ground-collision avoidance, early use of calculation machines in WW2 aircraft. Computational aeroacoustics at QMUL. Abraham Wald and the ‘red dots’, safety features, checklists, expansion of aviation safety into other fields, semiotics (safety information) and signage at airports (also an example of abstraction), airport semaphore comms by ramp agents. Aviation influencers on TikTok (computing as a sharing medium).

 

 

 

A couple of things I do when I apply for grants

1. Take the URL / link / address of the page the call is published on and stick it into Twitter etc to see what people are saying about the call and who is doing the saying. 

When applying for a grant a few years ago this helped me discover an interview with two of the people involved in the grant we were applying for. They were being asked what they were looking for and promoting the scheme. This interview was incredibly helpful but was not particularly flagged up anywhere and I pretty much found it by accident. 

Often, but not always, the first people to share the link will be involved in the project or organisation in some way so may have useful things to say about it.

2. See who has been funded previously and the type of project they did. I might search for the name of the grant on Twitter etc and add in words like "delighted" or "announce" etc.

Clicking on the link in their post or the website link on their profile can give me more info about that project but searching on Google can also put that information in context. This strategy recently pointed me towards an evaluation of a previous instance of the grant (because the organisation was listed in the evaluation), which also pointed me to more funded project examples and more info about what the funders had been looking for. 

Commentary
I think of this background research also as 'chaining' or 'triangulation' - using one bit of information to find another, then using that to search for other stuff and so on. Possibly everyone is already doing this and so I'm not doing anything unusual, but I've not seen it explicitly mentioned (or failed to notice because I'm newer to grant applications*) so thought I'd write it down here in case it's useful.

Background
Although I've worked in academia since 2010 it's only more recently that I've been involved in applying for grants. I've previously applied (successfully) for local small grants up to £10k and been on a few decision-making grant panels at that sort of level of funding.

More recently I've helped on some larger grants and am currently waiting to hear the outcome of one and planning an application for another mid-range grant when it opens again next year.

I've just noticed the 'teal and orange'-ness of some AI images & realised films might be to 'blame' (training data)

I googled to find something about the highest helicopter flown and read this article ('How high can helicopters fly?') which includes an image that I'm certain is AI generated (Ctrl+F as explained to jump straight to it)*. It's very "teal and orange" and I was suddenly struck by how many other AI images I'd subliminally noticed that also have this colour arrangement, though I'd not really elevated the notion to conscious thought until now.

Back in 2010 my friend @jjsanderson shared on Twitter an article called 'Teal and Orange - Hollywood, Please Stop the Madness' (March 2010) which explained the reasoning behind and overuse of these colour tones in film. It introduces an interesting artificiality to film which is visually arresting, but sometimes a bit too much.

Briefly: human skin tones are in the orange spectrum (pink to brown) and colour theory puts blue as the 'opposite' so enhancing orange and blue tones brings out skin tones and shadows in contrast. 

Squares of teal (left) and orange (right)

Wondered if others had noticed this appearing in AI images - yes. A year ago ;)

This color scheme shouts that your image was AI-generated
(April 2024 or March 2024 - date given as 03/04/2024)
This specifically references film as possible training data and also suggests, for those who wish to generate their own AI slop on Midjourney etc, some ways of getting round the 'teal/orange bias in AI generated images'. (It also pointed me to a TV Tropes article, Orange/Blue Contrast (date not given), which gives further examples of teal/orange use in film and television).

Here's a Reddit post asking the same question: Anyone else notice that some AI images look sort of yellow? (April 2025)

I'm sure there are plenty of other examples.

*Ctrl+F is a keyboard shortcut for Edit / Find on a computer. On a Mac it's Command F (press and hold the ⌘ then tap the F key).