@yacc143@mastodon.social

· 2mo ago

Replying to

@xgebi@hachyderm.io @EUCommission@ec.social-network.europa.eu Well, then you'll be happy that the enforcement is shared between MS and the EU Commission and it has become fully operational today. Responsible for that is the EU AI Office inside DG CONNECT. And for GPAI models (like large foundation models) the EU has exclusive authority, not the MS. (Full enforcement of these came online today, fines, mandatory evaluations, binding corrective measures.)

4

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@malteengeler@legal.social Aber ich vermute es ist besser über digitale Souveränität zu diskutieren als Teil des Managementteams eines Startup mit Multimillionären als Risikokapitalgeber Star als Anarchoaktivist ;) Das Lustige ist ja, mittlerweile hat ja die EU Gesetze die mehr oder weniger explizit digitale Souveränität einfordert. Also vielleicht nicht explizit, aber DORA macht im Prinzip Finanzinstitute haftbar für so Dinge wie den Cloud Act.

5

1

0

Andreas K @yacc143@mastodon.social

· 2mo ago

Replying to

@kasilas@mastodon.ie @xgebi@hachyderm.io @EUCommission@ec.social-network.europa.eu Well, generally the IE DPA has to be dragged into acting. By activists, Irish courts, or EU institutions like the EDPB. That's the trouble with MS-level enforcement; politicians literally see lenient enforcement of regulations as a competitive location advantage. Or to cite the initial Austrian implementation law for the GDPR, if my memory is correct, the DPA was actually instructed not to fine in regular cases, just to admonish.

2

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@johndgriffith@twit.social @DemocracyMattersALot@mstdn.social And that's the reason you won't get it, you would need the recipients to vote to disfranchise themselves.

3

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@bert_hubert@eupolicy.social You do realize that these side channel security issues in modern CPUs Spectre were initially researched by an Austrian tech university with EU grant money. So yes, it's maddening the EU did have surprising great tech, but what it lacks is the snake oil sellers willing to sell you every wheel as the greatest thing ever. Even if it's not very circular. While the Americans have no problem with selling you 💩 praising it over the moon.

3

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@michaelgemar@cosocial.ca The US won the first phase of the war. And then they made the mistake traditional American mistake of declaring the war prematurely as won. Without consideration of any of the fulfillment criteria for "declaring victory" were even checked. The Northern states had no real plan for after the civil war. Oops. Bush did declare Iraq won on a carrier of the US coast. And Trump declared Iran won without even having an idea how winning might be defined. @Deglassco@mastodon.social @benroyce@mastodon.social

4

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@malteengeler@legal.social Wenn dass dann schlagend wird sind das perfekte Gründe für viel juristische und wirtschaftliche Schmerzen und dann liest man auf LinkedIn dass es einen neuen CTO/CIO gibt oder was der Verein halt gerade in der Buchstabensuppe verwendet hat als zuständiges Opfer.

3

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@rl_dane@polymaths.social @virtualmarc@edgie.me @NanoRaptor@bitbang.social And yet I had an AI class in my CS curriculum in the 1990s. My point is that AI has always been a wide field, but as of the year 2026 it's a buzzword that has lost basically any information content. A ton of algorithms that were just algorithms a decade ago are today resold with a golden wrapper stamped "AI". It's been for some time the way to get i̵d̵i̵o̵t̵s̵ investors to part with their money. I mean that's why we don't market our solution as AI.

4

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@bigiain@aus.social Americans. We Europeans were so much simpler. We learned netiquette by lurking while mad downloading stuff from the alt.binaries... hierarchy. @davefischer@hachyderm.io @NanoRaptor@bitbang.social

2

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@johannab@cosocial.ca @pojntfx@mastodon.social @mitchellh@hachyderm.io I wouldn't go so far. But the system is changing. Some parts are accelerating. As the OP has mentioned, that might sound good on the surface. But there is always the law of unintended consequences. One of is that the faster "velocity" has the side effect of unmasking many issues that were there too before, but less of an issue when the red button "self destruct the company" was only accessible to slow working human employees,

4

5

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

OpenAI claims that it's clever enough to spot when it should switch to the better "thinking" LLM. Two problems with this:

it means the idiot brother decides if he needs to call in his more clever sister to solve a problem. No concern here, right? Or go on and use his mental hammer on the glass door. Or the suicidal user.
and the free tier has basically no quota of thinking tokens.

So quite often it's the lobotomized cousin that stands in for the LLM industry in all the horror media stories

3

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

who considered things like do I want to become unemployed before trying the button out. Now the same permission setup that was completely broken last year and a fatal risk, connected to via MCP server, can be triggered inside 24 hours by a stochastic NLP application. Because that NLP application does not consider things like if it wants to be employed tomorrow, but if you prompt it with “test everything,” who knows how it will interpret it?

3

4

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@coreyspowell@mastodon.social @MichaelPorter@ottawa.place touche

3

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

the basket, the A7 in the EU treaties would have allowed them to deal with Hungary relatively strongly. They did not, mainly because that would have created a precedent that the EU can and does intervene with the internal political processes of the members, and that's their own skin (the MS own political class and governments). They let Hungarians endure Orbán's regime for over a decade. Until they managed to vote him out themselves.

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

What I often see missing in discussions about AI (especially the current LLM centered discussions, AI is way more to be honest) is the context analysis:

Oh, oops, see what bad things happened because of AI.
Yeah, right, the AI might be an enabler, but if you behaved like that with a human instead of an AI, the outcome would have been as bad. Risky behavior/business processes don't become safe just because you trust some random dude not to be a serial killer.

2

3

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

To put it differently, if you trust things that some random dude says without checking/validating them, my answer to that would be “They eat the dogs; they eat the cats.” I've got news for you: humans lie. It's a human survival strategy that works. People lie on their timesheets. The GOP even acknowledges that they lie because it works in the polls. But some random machine that you know to have issues with that—you expect it to be telling the truth by default? What is even the truth?

2

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

So yeah, LLM are magnificient (compared to what we had before) NLP algorithms. But using them wrongly, and you end up with troubles. And the companies running the circus don't make it easier either, e.g. OpenAI's ChatGPT instant mode (which is basically the only thing accessible on the free tier) is literally tuned to provide fast, convenient, cheap-to-generate answers, that nobody cares if they are wrong. See what might be wrong with this?

2

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@Tutanota@mastodon.social The issue here is that the meme is misleading. A surprising number of MS are in favor of mass surveillance. It's just that it might not be the right thing to say in public, so you have to camouflage it behind a heroic topic, e.g. "protect the kids".

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

According to official data, company registry is public, sigh, he's only a board member as of April 2025. https://www.ratsit.se/5592384001-Mullvad_VPN_AB

ratsit.se

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@stormii@darmstadt.social I hadn't had my morning coffee yet, so I'll skip on giving it the other numbers but this rings like it hits some OWASP LLM items. But how exactly did they arrive at the idea that it would be a great idea to automate account recovery, via LLMs? If they insist it has to be done by the book, implement a classical process. If it's an ad-hoc decision, use the classical ad-hoc processing machine, homo sapiens. @briankrebs@infosec.exchange

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@icing@chaos.social Who could you be taking about. Whistle. Not that shoddy coding agent that is one hand extremely free with tokens and OTOH extremely pinchy when it hurts.

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@mkreutzfeldt@mastodon.social Habeck und die Grünen, natürlich, oder? Als loyaler Unionspolitiker würde er niemals einen Vorgesetzten anpatzen.

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@mcnado@mstdn.social Well, you know, Jesus was a woke communist. That's why the good GOP governor crucified him. Didn't you learn anything in Sunday school.

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

Anders formuliert, das Thema "Firma löschen wegen falscher Permissions" (Keywords: Single AWS Account, AdminAcces, IAM Keys, Backups im selben Blast Radius) kann ich mich bestens erinnern aus meinem vorheringen Startup Abenteuer (datieren wir es vorsichtig 2013-2015) und da waren AI Agents nicht das Problem, sondern menschliche Akteure. Das Risiko „ein einzelner Mensch mit den richtigen AWS-Rechten kann die Firma praktisch löschen“ war 2013–2015 definitiv Thema.

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

Entsprechend komme ich zur meiner Beobachtung, OWASP LLM hin oder her, es kommt immer auf Prozessdesign an, und Risikoanalysen. Wenn dein Mitarbeiter (Homo sapiens) die Firma in die Luft jagen kann (also virtuell), dann wird der LLM Agent, der potenziell zu viele Rechte für sein Herrchen ausübt, definitiv die Firma in die Luft jagen. Eine reine Zeitfrage, statistisch gesehen.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@rl_dane@polymaths.social @virtualmarc@edgie.me @NanoRaptor@bitbang.social Not really. Considering that AI has became a buzzword for anything nowadays, who knows what "AI but not yet called AI" was used in 2013 to creat them?

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@jim@social.openrightsgroup.org It's not “witchcraft” (nice typo, btw), but German experience with the Nazis and their abuse of databases (back then mostly in paper form, but they already used US-provided Hollerith systems too) for prosecuting minorities. Oh right, the American right does not consider the Nazis necessarily the bad guys anymore, isn't it so? And the Peter Thiel faction. #USPolitics #depol

0

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@womble@infosec.exchange Transparency is good, isn't it? Now if the gals and guys had slightly nicer hobbies, they might appreciate transparency about their doing more. @GossiTheDog@cyberplace.social

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

Logically, in an ideal world, the leaders of all the others MS would have seen that their duty was to protect the democratic institutions of fellow EU-citizens. They considered their "duty" to protect their power and the "sovereignty" of their own member states. Coming back to the US, I think post-Civil War there is a tendency to whitewash all the compromises and harsh interest conflicts that went into the formation of the USA and how they were papered over.

0

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@HasnainKazim@mastodon.world Könntest du deine Reisepläne veröffentlichen, das würde es erleichtern dich mit Falschinformationen zu versorgen.

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@johndgriffith@twit.social @DemocracyMattersALot@mstdn.social I completely understand that one wants to think big. But politics is the game of the doable compromise (the compromises happen in two-party systems generally before the elections, in European systems after the elections). As an European example: the EU MS had a long period where they could have reigned in Orban in Hungary (before the Poland went over to the evil side too) while he destroyed democracy in Hungary. As long there was only one bad apple in

0

3

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@sglerman95814@mstdn.social Starting with the wrong assumption that the UK died have representative democracy. Where exactly is the representative in your system when practically every government formed in the past century was based on a minority of the voters? Rule of law? Parliament is sovereign. That kind of makes rule of law a gentlemen's agreement. Actually everything becomes such. And somehow the behaviour of the modern gentle beings is somehow lacking. @NotTheLBCGuy@mstdn.social

0

2

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@samlitzinger@journa.host Actually, #Trump & #GOP did some preliminary work to disable the checks and balances. OTOH, I think you have a mythological misunderstanding of the Founding Fathers; they were also bound by political realities and compromises that were necessary to create a federal entity.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@grep@social.bawnet.io @icing@chaos.social Not really, the automation in reality still needs human (and generally experienced) humans for supervision. Or put differently, an highend LLM tends to generate nice code reviews, but I generally only pass 20-30% of the comments on -> the rest either are bullshit (hallucinations, yes, they still do that, or overzealous conclusions because it does not have the full context …), or the LLM does not have the full architecture roadmap (so it wants to go backward in the big picture).

0

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@malteengeler@legal.social Witzig. Ah, jetzt habe ich es. In unserem Prospekt verwenden wir den Begriff "digitale Souveränität". Und sind ganz brav bei der Vorzeige Cloud der EU Kommission zu dem Thema. Geht eigentlich nicht anders, zwischen Schrems II und der Verarbeitung von A9 Daten (biometrische Daten), ist jedes Setup das nicht DORA (nein wir fallen nicht drunter aber unser Biometrie Stack ist bei näherem nachdenken mit einem Stammbaum aus der Ecke belastet ;) ) konform ist fahrlässig.

0

1

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

@alienghic Makes me wonder when crawlers start packing reCAPTCHA-grade ML for this. Clean RL problem: agent walks a honeypot URL, IP/ASN gets banned, eventually the correlation surfaces and the policy learns to avoid traps. Guess not enough sites do it yet for the signal to bite. Genuine q: is your fail2ban purely local, or does it share ASN/CIDR bans via a community pool? If per-site, a crawler's done with you before the ban bites — kind of CrowdSec's pitch over fail2ban.

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@SeanCasten@mastodon.social A very concise legal theory and political program, wouldn't you say. And 6 people is a surprisingly small unit to win the War of Northern Aggression, finally. Guess the USA did not really win the civil war.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@pluralistic@mamot.fr It's more urgent than many realize. Most popular IDEs have now switched for some time to use using LLM based code completion. In some cases it's rather insidious, e.g. VSC, where the switch happens when you log in to github (even a free account), in rather underhanded way, or as an team member of mine told me, it's hard to get rid of that stuff, it tends to resurface from the dead like a zombie. In the US the CRO demands an exact accounting of what is human/AI written. Have fun.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@heafnerj@mstdn.social @dangillmor@mastodon.social @mastodonmigration@mastodon.online You personally might have pulled in different directions, but the Americans as a whole still elected the wrecking crew.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@coreyspowell@mastodon.social But is it a map? It's full of billions years old stale data, isn't it? It isn't even a historical map, for that or should show the data at a certain moment in history for the whole map. Just being silly.

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@heafnerj@mstdn.social @dangillmor@mastodon.social @mastodonmigration@mastodon.online well you voted them in. Not only Trump, but have him Congress. After he delivered such a successful trial period in Trump 45. Plus, with the captured SCOTUS arguably the criminals are defining what is criminal so it's problematic calling them criminals, isn't it?

0

1

0

Andreas K @yacc143@mastodon.social

· 2mo ago

Ah the DEA says the Trump attacks on the drug boats didn't slow the flow of drugs.

Hypothesis: that would work perfectly with that finding, KISS, without suggesting that cartels magically rerouted their drugs overnight: the boats were really just fishing boats that the Trump regime did sink to demonstrate how mighty the US Navy is.

And how willing to commit war crimes.

If the boats didn't transport drugs then removing them wouldn't change the flow of drugs, would it?

0

Andreas K @yacc143@mastodon.social

· 3mo ago

Replying to

On "no official source": under Swedish law (Lag 2018:90 §27), Kammarkollegiet must redact the names of individual donors from the public register — so the 5M line shows, but never the name. That's by design, not a gap. The name only comes out via the unredacted filing or, as here, the donor confirming it himself. https://www.riksdagen.se/sv/dokument-och-lagar/dokument/svensk-forfattningssamling/lag-201890-om-insyn-i-finansiering-av-partier_sfs-2018-90/

riksdagen.se

Lag (2018:90) om insyn i finansiering av partier

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@chrisstoecker@mastodon.social Es findet halt eine Verschiebung. Auf der einen Seite eine gewisse Beschleunigung statt. Da LLM nun anders handeln als Menschen, muss man plötzlich sehr viel genauer Berechtigungen etc nachdenken -> ein zu breiter IAM Key kann natürlich auch bei Menschen zu gelöschten Firmen führen. Mit AI Agenten wird halt so ein Key schneller und öfter zu oopsies führen. Zur Erinnerung: https://www.theregister.com/software/2017/02/01/gitlabcom-melts-down-after-wrong-directory-deleted-backups-fail/1228321 Kriminelle Energie geht auch immer: https://crimeline.co.uk/wp-content/uploads/2019/09/needham2019ewcacrim1541.pdf Alles von vor 10 Jahren

theregister

GitLab.com melts down after wrong directory deleted, backups fail

Upstart said it had outgrown the cloud – now five out of five restore tools have failed

0

1

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

That does not change the fact that it generally quickly finds a couple of “truffles” that one can “complain” about or generally improve. — But the point of AI doing complex tasks without human supervision—we are still far from that point—and that's from somebody who's mostly AI positive.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@internetarchive@mastodon.archive.org But it's

so embarrassing when it archives our oops moments
it's a classical path to circumvent payment systems for news sites

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

@jim@social.openrightsgroup.org It's BTW the ugly historical argument behind the continental European “fetish” with privacy and the GDPR.

0

Andreas K @yacc143@mastodon.social

· 4mo ago

Replying to

keine KI involviert. Dann wäre die AWS S3 Outage von 2017 wo ein interner Mitarbeiter durch Fehleingabe S3 in East-1 sabotiert hatte. https://aws.amazon.com/de/message/41926/ Aber wie gesagt, wenn man einen roten Knopf (virtuell) hinstellt, und mit "Firma Self-Destruct" label versieht, werden viele Menschen einen Umweg um den Knopf machen, weil sie sich Gedanken über ihren Beschäftigungsstatus +24h machen. Ein LLM hat da diese Kontextualisierung nicht, und wenn es jemand "Räum alles auf" prompted, ...

Amazon Web Services, Inc.

Summary of the Amazon S3 Service Disruption in the Northern Virginia (US-EAST-1) Region

0

1

0