Notices by Nicolas Kayser-Bril (algowatching@social.nkb.fr)

All in all, you don't need to read his book to understand the military value of Lavender and Gospel. (But the absence of thinking behind these systems does explain a lot about the catastrophic failure of Oct. 7).

My hypothesis is that automated target generation does not help with strategic or tactical goals but provides a moral absolution to kill and destroy without accountability, just like that Milgram experiment.

end/ 🧵

🇸🇪 Meta let criminals run ads for a scam. Several Swedes fell for it and lost €21m. They're now suing not Meta, but EY, the auditor, for failing to audit Meta properly.

https://www.mynewsdesk.com/prio-startup/news/ernst-and-young-sued-for-sek-240-million-for-failing-to-audit-facebook-correctly-481788

This is another attempt to hold Meta accountable for the thousands of scams it facilitates.

A case in Australia, where the consumer protection authority is suing on grounds of abetting criminal activity, is still ongoing. (Here's my report from 2022: https://algorithmwatch.org/en/meta-sued-for-abetting-fraud/ )

It's been 7 years since an official report described the conditions for asylum seekers in Libya – largely encouraged and paid for by EU and Member-States – as "like a concentration camp".

https://www.faz.net/aktuell/politik/ausland/fluechtlingslager-in-libyen-diplomaten-warnen-vor-kz-aehnlichen-verhaeltnissen-14786149.html

Together with some excellent journalists (e.g. @mathildesaliou) we talked to a dozen people in Europe who were actually fighting back against AI systems.

Their stories are sobering: Seeking redress is always hard, rarely works... and not one mentioned the AI Act or the DSA.

https://algorithmwatch.org/en/europeans-push-back-against-rogue-ai/

An app-taxi company and the ad agency BETC ran an campaign in November claiming that MidJourney systematically misrepresented the French banlieue as shabby, sad and derelict. (They encouraged inhabitants of the suburbs to send postcards to MJ employees to fix it).

https://betc.com/en/society/heetch-greetings-from-la-banlieue

Someone on LinkedIn just pitched me this new "AI fact-checker". Works as intended.

Jokes aside, the Vatican is once again much better than most at AI policy. Tidbits:

- Contrary to what's reported (gotta check the pope's footnotes, folks), Francis did *not* buy into the existential risk discourse (the risk he meant was climate breakdown)

- "Algorithms do not offer guaranteed predictions of the future, but only statistical approximations"

- AI "should always be regarded as socio-technical systems"

🇫🇷 The French welfare agency uses an algorithm to detect potential fraudsters. Every month, it ranks the households of 32 million people, including 13 million children, with a "risk score".

For the first time, @LaQuadrature obtained its source code and could run it.

The criteria that increase the risk score are
- being poor,
- living in a poor neighborhood,
- being a single mother,
- being unemployed,
- having a disability.

https://www.laquadrature.net/2023/11/27/notation-des-allocataires-lindecence-des-pratiques-de-la-caf-desormais-indeniable/

🇪🇺 Art 22 GDPR: You can't be subjected to an automated decision with legal effects.

🇫🇷 French authorities: We processed your personal data through Google Bard and denied your asylum claim.

https://pastebin.com/ENFGeErT