2022 has shown that we must increasingly approach digital media with a certain mistrust (newsletters excluded).
So-called artificial intelligence (AI) can now generate texts that at first glance appear as if they were written by school children, journalists or scientists.
This is fascinating or entertaining until it becomes a problem: spam, fraud, misinformation or even faulty computer code - all of this is massively simplified by AI.
After all, experts have recognized this.
The development of suitable detection methods is ongoing, but above all it is lagging behind.
"The enormous speed of development in this sector means that any methods of recognizing AI-generated text very quickly look very old again," writes Technology Review.
It's a race, "and at the moment it looks like we're going to lose."
And that's just the beginning.
In photography, we're close to being able to generate images in seconds that show things that never happened or people that were never in the same room.
This may be extremely practical if you no longer have to buy stock photos, but can simply create them yourself.
But of course this technology is already being abused.
The near future may be even wilder.
The Israeli newspaper Haaretz has published a report suggesting that even surveillance cameras do not always show real events.
Co-founded by Ehud Barak in 2018, Toka, with offices in Tel Aviv and Washington, writes Haaretz, selling technology that allows its customers to "locate, hack, view, and manipulate security cameras or even webcams in the area." , including earlier recordings«.
The whole thing should be possible almost without a trace.
Barak is not only the country's former prime minister, he is also considered one of the most decorated soldiers in Israel.
A company presentation quoted by the newspaper specifically states that one could "manipulate visual recordings 'to mask on-site activities'".
In other words, secret service or military operations may remain invisible because a hacked surveillance camera shows something completely different.
Haaretz is reminiscent of the 2001 film Ocean's Eleven, in which a gang of crooks hacked the surveillance system of a Las Vegas casino and played fake recordings.
This is exactly what is possible with Toka's technology.
Hacking cameras might be the easy part
The article does not describe exactly how this should be done.
I think hijacking cameras is the easy part: Many of these devices are likely to run with outdated firmware, so there are certainly plenty of weak points for specialists to find.
In a way, hackers already played out the scenario in 2015, even in Las Vegas, albeit only officially on the stage of the world's largest hacker conference, Defcon.
For me, there are more question marks behind the image manipulation: Can the transmission only be redirected to another image or signal?
Are shots being altered at the pixel level with the help of artificial intelligence?
How fast is that?
Or is that exaggerated self-promotion?
As expected, there is nothing on the company's website, there is no mention of hacked cameras at all, only of "remaining undetected to maximize operational efficiency".
To me, that's the most flowery paraphrase of the year for covert espionage operations.
According to this, Toka's customers include government and law enforcement agencies as well as secret services almost (!) exclusively in the West.
The company confirmed this to the newspaper and emphasized that it only works with the United States and its allies.
There is also a "rigorous" approval process that respects "law and order, signs of corruption and civil liberties."
Under no circumstances do you sell your products to private customers or individuals, or to countries and entities that are on the US sanctions list or to which you cannot ship anything from Israel.
"Less than a fifth of all countries in the world" remained.
The documents available to Haaretz show - albeit without further details - that the company has already made representations in Germany.
For that reason alone, I believe we'll be hearing and reading even more about Toka in 2023.
And that we must learn to doubt images from surveillance cameras.
In any case, for anyone working with digital evidence, the future will be more exciting than any »Ocean's« film.
Our current Netzwelt reading tips for SPIEGEL.de
»This is how German hackers got hold of US military biometric devices on Ebay« (six minutes of reading)
From Afghanistan to US dealers to Matthias Marx: How a CCC hacker was able to buy six biometric devices used by the military and whose data he found on them, he has Max Hoppenstedt told.
»Does the EU want to regulate Rezo's videos?« (five minutes of reading)
Brussels is planning restrictions on the targeted dissemination of political advertising on the Internet.
YouTubers fear an encroachment on freedom of expression, Torsten Kleinz agrees.
"When the computer plans the Christmas menu" (four minutes of reading)
SPIEGEL-Netzwelt had three recipes written by GPT-3 and copied them.
Now we know how artificial intelligence could one day take over the world.
External links: Three tips from other media
"Save energy?
Why switching off the router has more disadvantages than advantages« (four minutes of reading)
The colleagues from »heise online« have calculated whether it is worth switching off the router at night.
The result is interesting but sobering.
»Welcome to Video« (Podcast, English, 68 minutes)
The famous US journalist Andy Greenberg has written a book about investigative methods in the world of cryptocurrencies.
The podcast "Darknet Diaries" is about a case that Greenberg describes in the book.
But be warned, this is about child sexual abuse.
The episode is very interesting, but it's no fun.
"Man simulates time travel thanks to Stable Diffusion image synthesis" (English, two-minute read)
What if we could travel back in time, to Troy, to the building of the pyramids, or to the time of the mammoths?
An artist shows it using AI-generated selfies (of a non-existent person).
I wish you a good start into the new year,
Patrick Beuth
