“Ministers embraced the ‘world’s safest phone’. Then it unravelled” – an article by Georgina Quach in the Financial Times. She writes,
Rather than relying on apps to police harmful content, governments now want to intervene at the operating system level, so guardrails can get ahead of whichever app breaks on to the scene.
That is why Keir Starmer in June presented an ultimatum to tech companies: install child nudity-blocking software in devices by September or face legislation and fines.
And there is no excuse for dithering, the government’s announcement implied. It said this tech had already been achieved by the British company SafeToNet. Ministers held up SafeToNet as proof point for the proposition that child safety can be built into phones universally and block nudity in real time.
But what appeared to the government to be an “oven-ready” fix is now running into difficulty, after the first smartphone to integrate SafeToNet technology was removed from sale following an investigation into the device’s safety risks. The HMD Fuse, sold by Vodafone, was launched last year as the “first smartphone that protects a child’s innocence by stopping nude content from being filmed, seen, shared and stored”. In January of this year, former ministers Peter Kyle and Jess Phillips starred in a video promoting the phone. Six months later, sales have been suspended.
Why? The answer is alarming:
Then, in July 2026, Moore discovered what he described as a different, serious flaw: an unauthorised person could potentially register as the guardian of another child’s handset and access its live location. According to Moore, the problem was that the pairing relied on the IMEI number, the unique serial code that identifies each handset. He said this number was predictable: part of the number was fixed and the rest could be systematically enumerated and plugged in. He said that meant he could “become the guardian of every child on the platform, which meant I basically had a live map of every child on the system at the time”.
I just asked chatgpt whether this policy fiasco could have been predicted based on the precedent of the promotion of compact fluorescent lightbulbs vs light emitting diode technology. The output was a bit on the verbose side, but it would not have required much editing to be quite at home posted above-the-line.
It gets even better. SafeToNet is basically a bunch of guys in a an office with no serious tech background. Just a some guys with very limited IT experience mainly in the City. Their VP Engineering is some IT guy (not tech guy) whose main business seems to be “life coaching” business executives.
They did raise about 35 million quid, have peanuts revenue, and losses equal about 50% of what little revenue they have. Their “AI Tech” seem to be pretty much duct tape, bailing wire and the sort of software you would see written as part of some (bright) undergrads end of term project. They dont seem to have hired anyone with any serious software experience.
The IMEI fiasco is a great example of their technical incompetence. If you want to pair robustly you use could something like SSL/TLS certs. Or even device profiling to generate a unique device fingerprint. A bit like how copy protection on PC software has worked for almost 50 years. This is Security 101 stuff.
In other words they are charlatans.
Anarcho-Tyranny continues.
Government regulations are arbitrary – and certainly do NOT protect children.
Indeed the same government (elected and unelected) that pretends it wants to protect children from sexual content – pushes what used, up till very recently, be considered sexual perversions on children – via the education and the media.
As Prime Minister Gladstone used to say – of one thing I am certain, it will not be via the state that the morality of the people is improved.
Rather the contrary.
As for the real aim of the internet regulations – pushing political control and crushing dissent.
It was ever thus – after all that was the real aim of Prime Minister Walpole’s theatre regulations some 300 years ago, the propaganda that the aim was to protect public morals – being a vicious lie. The play he showed the House of Commons, to horrify them into censorship, a play filled with murder and rape, had been paid for by Walpole himself – his real aim was to stop plays that attacked the government.
Specify “nude” in a computable manner.
https://xkcd.com/1425/
jgh: That’s a bit behind the times, I think. A GAN running on the device could probably do a reasonable job. False positives and negatives might be a problem, but you’d be surprised at how good these things are nowadays.
You’d need a heart of stone not to laugh
It reminds me of Wiley Coyote and the Road Runner. Always some brilliant scheme to catch him, always falling flat on their face.
And I might add, despite the constant application of utterly fatal mistakes always seeming to get up and try again the same shit the next day.
@Sam Duncan
jgh: That’s a bit behind the times, I think. A GAN running on the device could probably do a reasonable job. False positives and negatives might be a problem, but you’d be surprised at how good these things are nowadays.
Indeed, in reference to the xkcd cartoon, Google Lens is excellent at identifying birds.
Having said that, running a sufficiently powerful GAN on most phones in a continuous monitoring mode would take up a huge amount of the phone’s computational and battery capacity. If you attached it to a live stream from a video camera I think its impact would be quite significant. I guess they could put in a special chip just for that, but those chips are not cheap and it definitely would impact the battery life.
I see it as the God delusion all politicians seem to have.
“I have come up with this brilliant idea.I have issued a commandment. Now do it, mortals”.
Having the square root of zero comprehension of how technology or the world operates, how long it might potentially take or any difficulties involved, then it ALWAYS works out being a disaster of one kind or another.
We need to return to the root of it all: Woodrow Wilson’s governance by experts ideas. We need to destroy his ideas! A public trial condemning those ideas, the people who espoused them, the idiots who follow them not knowing the origins, and all laws, rules, customs, regulations, bylaws, that have been issued because of them.
Here’s a serious question (I think…I hesitate to label anything I say here as serious given how much smarter so many here are): how do we achieve a comprehensive rollback of the evil of Wilson’s ideas?
And by “rollback” I don’t mean a repudiation by logicians–no one cares what we write here–but a political revolution against the Administrative State.
I actually think a nominally un-related thing could help: legislate against the collection and sharing of private data. This would upend the modern “tech” economy (sneer quotes as what most urbanites think of as tech is anything but…it’s just software, and usually very bad bloatware at that!). It would de-monetize just about every “free” app you have (more sneer quotes…those apps aren’t free…this proposal would make you pay for them up front vs in the hidden ways you now pay).
It would cost people a lot, but mostly it would cost the people who can most afford it. The guy who drives a 15-20 year old car with no smarts would not be affected. He uses his “smart phone” for phone calls and texting. His “betters” use theirs for stock tracking, social media trend analysis and god knows what else. That would all get screwed. While this nominally would affect tech and significant parts of the economy, it would strangle the deep states data pipeline! Win-win.
less screen time for all (he “says” while typing and looking at a screen!)
@GregWA I’m not sure I understand the connection between the foolish notions of State micro-mangement and the idea of free software services exchanged for tracking data to facilitate advertising. Fundamentally the problem with the state isn’t really so much the “experts” who do things so much as its pervasive reach into every corner of our lives. I think, for example, insofar as we have to have things like building codes and road design and even centralized school curriculum design (or at least minimum standards of education) then we do need experts to design them. Insofar as we have a military and police force and courts of law we must surely have experts designing and running them. Sometimes these “experts” are more political activists rather than real experts, but that happens most commonly in areas the state should never be involved in.
Again the problem is not that we have experts in the government, it is that the government is far too large and deals in areas where private experts, operating in a competitive environment would be vastly more effective.
As to free mobile applications, I think it is actually the poor who benefit most from them. Personal data from poorer people is of the least value, since it is almost entirely collected to drive targeted advertising, and the poor by definition are the least attractive target group for advertisers, whereas they get immensely valuable services at not cost. And often times they have a choice in some applications between a free version where they give up information and time-to-watch-ads, or paying a small fee and skipping that. This seems to me the free market at its very best.
The guy with the old car benefits greatly from this, both he no doubt uses google maps to find his way, youtube for videos on how to fix his clunker truck, Gas Buddy to find the cheapest gas, and free music to entertain that is much less crappy than listening to the radio. And of course a vast source of information and news from diverse sources rather than the curated, managed politicized information that we all had before the Internet burst that particular bubble.
The Law of Unintended Consequences, a continuing series. The next edition will start after the weather, sport and travel. Enjoy your weekend!
Fraser, I agree…it is a tenuous connection at best. But you said it yourself: “Fundamentally the problem with the state isn’t really so much the “experts” who do things so much as its pervasive reach into every corner of our lives.” It’s that reach that I want to roll back by making data more private, or really, putting control of our data firmly in our hands.
As I’ve thought about this, I’m not sure there’s an implementation where the benefits I perceive outweigh the costs to a LOT of people. I guess that was my other thought: what cost is worth a significant rollback of data access of the Administrative State’s experts?
“The output was a bit on the verbose side”
Isn’t that a bit redundant? Everything from chatgpt is on verbose side.
Sam & Fraser: After reading up a bit on GAN, it still needs a data set to be trained on, so needs a supply of known nuddy pics. Supply a source of known pictures of birds tagged with the known identifying data is easier than the murky task (noth technically and ethically) or supplying a source of pictures of naked people, and identifying the correct “nakedity” of them. Christ on the cross to be banned?
@jgh
supplying a source of pictures of naked people, and identifying the correct “nakedity” of them. Christ on the cross to be banned?
I think you need to understand how these data structures are generated a bit better. It is a complex interaction of raw data and human testing and tagging, and often what they call RL training, which I think is ironic since RL usually means “real life” whereas in this context it means exactly the opposite — training with synthetically generated data (it is an abbreviation for “reinforcement learning). So, for example, they might supply a bunch of photos of Christ on the cross and say “This is ok” and that data will be incorporated into the neural net. One thing I do wonder is how they can get training data since CSAM images, even if used for legitimate law enforcement purposes are absolutely radioactive, and not only as raw data but it would have to be reviewed and tagged by humans. So I guess the government would have to do it, and these idiots couldn’t organize a piss up in a brewery.
FWIW, there was a famous USSC case where Justice Potter said he couldn’t define porn but “he’d know it if he saw it.” This is a fascinating statement in this context because it illustrates exactly the difference between traditional algorithms for this and AI ones. With traditional algorithmic approaches one has to work out a precise definition of what we are looking for, what exactly is porn. But AI is quite the opposite, here you say “here is a lot of what porn looks like and here is a lot of things that aren’t porn” and it works it out from there through the training process. Which is to say it knows whether a picture is pornographic because it’ll know it when it sees it, just as Justice Potter said. Think of it like saying to the computer — here are one billion porn images, and here an one hundred million images that might be considered porn but aren’t. Now look at this picture: is it similar to one of the first group but not one of the second group?
Once the network weights are generated they are loaded onto the computer as a large matrix of small numbers for the inference engine, and so the source pictures cannot be reverse engineered out of it. Though, again, with the idiots from the government, I wouldn’t put it past them to screw that up too.
That’s what I was saying. You need to feed it loads of pictures of nuddy children that some human somewhere has looked at and tagged as nuddy children. And consistantly. How do you ensure that all the humans doing the tagging consistantly tag the same? Is Airplane porn ‘cos it’s got nuddy bits in it? Are mentally disturbed men in frocks waving their willies around in public porn? Is Nirvana’s album cover porn?
It was never about protecting children.
The same establishment that pretends to want to protect children from sex on the internet, pushes sex on children via the education system and media – especially what used to be called sexual perversions.
The internet controls were always about political and cultural control – children were just used as an excuse.