FTC proposes new protections to combat AI impersonation of individuals
ftc.gov
ftc.gov
> Falsely imply government or business affiliation by using terms that are known to be affiliated with a government agency or business (e.g., stating “I’m calling from the Clerk’s Office” to falsely imply affiliation with a court of law).
My first thought on this was great, I’m glad the FTC is doing something about this and I’m surprised it wasn’t already regulated by the FTC.
My second thought was the majority of this type of fraud is probably from foreign impersonation, not Americans. And it’s not like they’d be sending in predator drones for surgical strikes against scam call centers, as satisfying as that might be.
My third thought was that having this on the books and keeping a record of these violations will give the FTC leverage to crack down on telecoms that don’t do anything about it.
There's an industry for it in India, because they can source english call center workers there fairly easily.
The youtuber Mark Rober ran into a couple of US mules for these sorts of scams while working on one of his prank videos, and decided to team up with some other youtubers to troll some of these companies in India. The video on it is worth watching: https://www.youtube.com/watch?v=xsLJZyih3Ac
I think it means that american-owned social media platforms would be required to comply with these rules though, right? that seems like a fairly big deal.
robocalls should just be illegal. unidentified propaganda should be banned. ads with fraudulent claims should be prosecuted.
you can argue about "to what degree" but AI isn't doing anything but exposing the true lack of enforcing of existing laws because of capitalism grease.
Ban robocalls, though, and just pay real humans to make those calls instead of overpaying superintendents.
As we know, a weakness in all free-entry point systems (like email) is freeloading. The somewhat broken "solution" is charging a small fixed fee to the initiator. But this doesn't work in systems we would like to keep at zero cost because it's a blanket response to an acute problem - most coms are welcome and legitimate.
Giving the callee the power to charge the caller a small "handling charge" if they don't like being bothered makes a lot of sense.
Of course it has its own modes of abuse, but the best way to avoid that is not to send to people you don't know.
It would also be nice if they streamlined the ability to add yourself to the DNC list.
Google engineers that are here, is there a reason you don't do this? If not, can you do it in your 20% time? 350 million of us will thank you.
You used to be able to collect direct from spammers if you could hunt them down, too. I think CAN SPAM ended that, which is a shame.
I would go as far to say unsolicited business calls, automated or otherwise, should be banned.
This is a nice little resource for quick reporting that an HNer put together recently. I have the abuse report email templated now, so all it takes is 2 minutes to report every new one.
https://reportphonespam.org/#Reporting-abuse-to-your-wireles...
Shout out to Telnyx staff for actually processing my requests with a reasonable SLA. Your CS team should be the industry norm.
I would pay extra to be part of a phone network that guaranteed humans only and displayed First + Last Name & Organization (if any) of the caller.
Oh man, this sounds bad. Who gets to decide that something is the real, real truth and not propaganda? The US? China? Germany? a PAC? A company? Your neighbor?
Unsolicited and unidentified should be banned, with users being able to opt-in to any lowered bar (which would count as solicited) if they like.
If businesses or political actors start sending robots to shout messages into your home from the street, without some opt-in, they would be banned.
I would be happy for all unsolicited messages (mail, email, text, voice) without a specific purpose related to a relationship or related to unavoidable impact on the receiver be outlawed.
Citizens that were ok with unsolicited or unidentified advertising or political messages could opt into the mediums they were ok receiving such messages on.
This is just a dismissive way of saying “AI isn’t doing anything but making the problem worse.”
I guess my opinion is "they already weren't tracking the type of fraud AI is capable; they weren't enforcing fraudulent activity..."
AI just makes it more numerous. a non story.
"sorry guys, we're still not regulating white color fraud"
white-collar fraud
§ 461.5 Means and Instrumentalities: Provision of Goods or Services for Unlawful Impersonation Prohibited.
It is a violation of this part, and an unfair or deceptive act or practice to provide goods or services with knowledge or reason to know that those goods or services will be used to:
(a) materially and falsely pose as, directly or by implication, a government entity or officer thereof, a business or officer thereof, or an individual, in or affecting commerce as commerce is defined in the Federal Trade Commission Act (15 U.S.C. 44); or*
(b) materially misrepresent, directly or by implication, affiliation with, including endorsement or sponsorship by, a government entity or officer thereof, a business or officer thereof, or an individual, in or affecting commerce as commerce is defined in the Federal Trade Commission Act (15 U.S.C. 44).*
It's the "with knowledge or reason to know" clause that's key here. Various industry parties have already commented on this, some wanting stronger language there to protect sellers of general purpose tools for creating content.
Sellers of automated outbound marketing tools which can be used to deliver impersonation scams might be caught up by this.
What needs to happen is that the liability should be clear and present danger when a reasonable tool provider knows X is a spammer, and they don't investigate (or take action) the cost of maintaining that customer should be many times greater than the revenue from it. This should be particularly true if a spammer is reported by a consumer (Which is a strong signal).
Right now, the situation is reversed.
In the novel, despite several attempts from different parties to encourage people to think more critically about what they read on the internet, highly sensational AI generated content radicalizes and stupefies the population.
First, it's awful that this could be even considered as needed in the future, and developers behind the open source projects should consider the future they are enabling. It's not all just harmless tech.
So with that in mind, I've spoken to people who have theorized about releasing preemptive deepfake porn of themselves. This is due to the currently awful but very much existing trend of revenge porn, and possible expansions of the theme that w/ deepfakes. Can't blackmail someone via embarrassment if it is all plausibly deniable.
For example, I was in a con with a presenter who was a leading advocate against revenge porn, and the the call got zoom-bombed. Awful stuff. I could see someone on the receiving end of that treatment going full nuclear in the manner I described, in the off chance those sort of measures ended the threat finally.
There are a lot of 'reputation management services' whose job is to flood out bad press and replace it with anything else.
Artists don’t want their art style copied.
Writers don’t want to be out of business or have the price of their work degraded by GPT spam.
There are a lot of things people don’t want AI to do, but I can’t want for them to use AI to remaster Star Trek Next Gen (and other old sci-fi) into 4K. This kind of application hurts nobody.
I've seen a few "4K" AI-upscaled movies and they look horrible. Rubber faces and uncanny valley effects every few seconds. What a shame. I'll stick with 1080P on blu-ray.
I think the AI adds crispness that was not captured due to limitations in the cameras (and skill of the cameramen) at the time. Possibly it also fakes textures to make them more visually appealing.
If the remaster only improves video and audio fidelity sure. But they may also swap out characters, personalities, and plot points, based on the whims of the studio.
AI controlled by major powerbrokers is going to further the aims of those power brokers.
Civil is a much lower bar than criminal when it comes to court cases.
However, the false belief that a proxy persona, such as remote telephone, video call, or any other non in-person interaction are what give rise to impersonation.
Might as well simply say that all persona non-gratis communications are unenforcable.
Lawn mower manufacturers said they couldn't make lawnmowers safe, it was impossible. Until the government mandated that they had too.
We have insanely high quality printers, yet we do not have much counterfeiting.
Just because we can do something easily and illicitly, doesn't mean that people will do things illicitly if the proper instinctive structure is implemented.
We are talking about a software based solution that can emulate any public figure (locally) that the average person will not be able to recognize. This is a categorical risk to the information age
An easier requirement from official sources to embed a public key in a video. This is a solved problem
laws that say murder is illegal are not regulating knives.
Basically, you're suggesting a technological approach that's fine for a narrow set of concerns, but you still have to regulate and disallow behaviors that can be prosecuted. The actual problem here is not a solved problem, but yes in one very narrow subset of problems, a more complicated solution than hosting official videos on official sources would be to require official sources to also sign the video.
What concerns me is the asymmetric threat such as posed by North Korean dollar groups. For the uninitiated, this is a real threat but let's for a moment not think about the silliness of present day Communists getting extremely ruthless about stacking cash.
The model I'm concerned about is, say you have a NK hacker group and they make a very, very convincing video of a CEO doing something embarrassing (shout out a former UK PM's alleged porcine initiation ritual) with a view to making cash.
These people are focussed on the bottom line. They could structure their extortion demand to be F.O. money and get paid with little fuss. And do it over and over.
On the one hand the replicability of such attacks concerns me. However I have been considering a future where we are embarrassed or exposed to embarrassing content on an industrial scale.
Embarrassment is a social concept that we all deal with, and deal with it we do. It could be that the AI impersonation mess gets so bad we all become inoculated to this type of content because virtually everyone notable has become a victim already. Could it become the cost of doing business?
No it isn’t symmetric.
There is no balance between party A being fraudulent so party B now needs to be fraudulent.
In both cases, the public C is the victim.
A permissive tit for tat view of fraud, simply encourages an arms race of victimization.
Organized crime groups compete too. That competition doesn’t result in some kind of optimal societal impact balance in the absence of legal responses either.
It is still an open question but I think money will be dumped into this area, for protection of IPs and compliance
I've been working a bunch on neural networks, and I'm quite convinced that this isn't possible (as in, not that we haven't built a tool for this yet, but that it's not possible to build such a tool and weights can always be redone) - I'm not wholly certain as I have been surprised before, but I'd need to see some evidence to even consider this as plausible. Fine-tuning can change a lot, adversarial examples can target very specific aspects of the model.
Furthermore, people do train models from scratch - even if 99.9% of models would include unremovable watermarks, all fraudsters need is for a single unmarked model to exist; and if it doesn't, they can fund the GPU-time to train one since they're running a high-revenue business, fraud has more money than academia.