The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
The EU AI Act
Cybertoonz explains EU tech law in four panels
As covered in this week's Cyberlaw Podcast, the AI Act is getting some poor reviews -- from the US Congress as well as Europe's tech sector. How could something like this happen in the home of the vaunted "Brussels Effect?" Fear not. Cybertoonz has the explanation.
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
Oh no! The companies that spend lots of money on AI don't like AI regulation! And neither do the politicians that are funded by those companies! Whatever will we do!
I think the larger, more long term concern is not who backs, or does not back this or that regulation; its at what point does AI reach to level of sophistication that it develops its own agenda regardless of funding and will we even recognize it when it does?
I know that sounds like dystopian scifi fantasy but I really do think we are playing with fire here.
Gorsh! Sounds like someone who never met a bureaucrat he didn't like.
Q. What makes you think the government bureaucrats, who know nothing of the technology, can manage AI better?
A. Government, baby!
What makes you think that the officials who prepared and drafted this legislation know less about the technology than the CEOs of big tech companies with their MBAs? (Or their lobbyists)
Because if those officials knew anything about the technology they'd be working in the tech sector rather than for the government.
Ehhh, I agree on average, but there are smart people incentivized by things other than money.
Yeah, power and growing their fiefdom.
You sure have a dark view of humanity.
No, I have great faith in individuals, it is you statists who distrust them and think a government paycheck somehow imbues bureaucrats with infinite wisdom and knowledge.
Anti-statists prefer to replace a system of representative government, however flawed, with corporate feudalism, because we've seen all the things corporations like to do when they think they can get away with it.
So I never said anything about infinite wisdom. That's a helluva strawman.
You're the one coming in that only money and personal power matter.
Your great faith in individuals seems easily swept away when confronted with your inchoate hostility towards the public sector.
"dark view of humanity"
A realistic one, backed up by 5,000 years of recorded history.
We all know you are eager to declare everyone sucks as much as you do, Bob.
Perhaps too eager.
We all know you are a bureaucrat so are eager to declare bureaucrats have the highest motives.
Perhaps too eager.
We all know bureaucracy is evil and the sooner we get back to kings beng deposed by usurpers and killing everyone loyal to the king to start ruling with no reference to anything that went before, including land ownership and various rights and laws, the better.
Bureaucrat are just people. Some are cool, some suck.
That's it; that's the whole thing.
As I said, Bob, you keep posting everyone sucks as much as you. But you post it at the drop of a hat. It's always felt performative to me, but maybe you're trying to convince yourself as much as anyone else.
" maybe you’re trying to convince yourself"
Oh, mind reading about motives is cool when you do it.
I need to keep posting it because fools never seem to understand it.
Eh, that’s just idle speculation, not load bearing on any thesis of mine.
... but if you have insights you'd like to share, you're in luck. The European Commission just published another consultation about AI this morning.
https://twitter.com/lewis_crofts/status/1744682119958331528/photo/1
Oh no! Politicians are worried that an AI chatbot might be used to generate speech and content with ideas they want to outlaw! Whatever will we do!
How dare they object the golden new age of deepfakes and fake news plus!
Yes, better to leave the fakery to the government.
What a great argument for swamping the world with shitty disinformation and fake sex vidoes!
What a great argument for foisting Fauci on the world!
Yes, we should have run up an LLM to sort out the pandemic.
Wank
Disaffected, antisocial, right-wing, anti-government cranks are among my favorite culture war casualties.
And no problem replacement is not already solving.
EU regulation of tech companies is just jealousy of the US because Europe has few tech success stories. Nokia was big 20 years ago but europe is a tiny player, all the big companies are American with even China a poor second.
Spotify
I didn't say they had no tech companies.
ASML
Adyen
Having an actual woman in this situation would probably have an effect on this macho posturing. As we can see, it is outside the scope of his imagination.
Adding a woman would just lead to the first agenda item becoming the color of the drapes, and should they be open or closed.
At a white, male, socially awkward, obsolete blog, can anyone be surprised that the images selected by these obsolete clingers are straight from the 1950s?
The images look like an animated version of Mad Men. The guy that's standing could be Don Draper making a pitch for Lucky Strike.
The Michigan Citizens Redistricting Commission, one of those efforts to strip redrawing districts from legislatures, which cures all ills,
https://www.michigan.gov/micrc , has one of the sweetest, diverse pictures you'll ever see.
It was just ordered to redraw districts for racistly lessing the voting power of African Americans.
For those of you for whom fascades are important. To quote a wise, if problematic man, "It's all a joke."
Aw poor baby, are the hitorically oppressed and disempowered being given a fraction of a fraction of the voting power used to keep them in line?
FW/PBS, Inc. v. City of Dallas, 493 U.S. 215 (decided January 9, 1990): “adult” businesses had Free Speech right to contest zoning ordinance that allowed denial of occupancy without procedural safeguards: applying standard from Freedman v. Maryland, 1965: municipality cannot allow “unbridled discretion”, must put time limit on decision, and must allow for prompt judicial review (for judicial review of some embarrassing City Council dereliction as to an adult business, see http://www.metnews.com/articles/2005/lshs010305.htm, or 22 Cal. Rptr. 3d 805)
United States v. Chicago, N.S. & M.R. Co., 288 U.S. 1 (decided January 9, 1933): ICC can’t suddenly construe regulation as requiring trolley car company to get its permission before issuing bonds, after years of not requiring it from other companies
White v. Pauly, 580 U.S. 73 (decided January 9, 2017): officer who shot and killed man who was pointing gun at him from inside house was not acting in violation of “clearly established” law and therefore was entitled to qualified immunity (reading the opinion one sees ugly behavior on both sides, arising from a report of road rage and drunk driving)
Chao v. Mallard Bay Drilling Co., 534 U.S. 235 (decided January 9, 2002): Coast Guard vessel in state territorial waters was subject to OSHA violation on matter not addressed by Coast Guard regulations (uninspected well leaked natural gas and spark caused explosion killing four)
Flood v. White, 531 U.S. 225 (decided January 9, 2001): granting habeas relief where state supreme court had held that defendant’s conduct (operating unlicensed hazardous waste facility) was not within scope of charged statute but nevertheless had refused to free him and had answered Court’s certified question that its holding as to the statute was not retroactive
Get tired of waiting for Josh?
Yes
Maybe Prof. Blackman is working on the big announcement that he has secured a spot on the faculty of a good law school.
Or maybe not, and it's just a bug in the autoscheduling program.
Let's not joke and keep it plausible. If it's something like that, it will be that he's going to be Trump's White House counsel.
He was working on a brief in the Trump case.
If the S/C adopts his reasoning, its going to be a sweet day to dump on all you Blackman haters.
.
So we love amicus briefs now?
Also, if Prof. Blackman has been manually posting his daily Supreme Court posts, that is one of the saddest things about him.
Whew!
I was jonesing on not having my daily hit of "Today In Supreme Court History" / captcrisis.
Mucho beaucoup!
Re: Chao v. Mallard Bay Drilling Co
capt c:
You wrote, "Coast Guard vessel in state territorial waters was subject to OSHA violation," but the case appears to be, " . . . an explosion on board Mallard Bay Drilling Rig 52, a (private) oil and gas exploration barge, (which) killed or injured several workers while the barge was drilling a well in Louisiana's territorial waters. The Coast Guard's subsequent investigation did not accuse Mallard of anything, but did note that the barge was not an 'inspected vessel' subject to comprehensive Coast Guard regulation.
The Occupational Safety and Health Administration (OSHA) then cited Mallard for violations of the Occupational Safety and Health Act (Act) of 1970." (oyez)
FYI, unanimous decision with Justice Scalia not partaking; doesn't list why he sat out.
Thanks!
Reports indicate it is not going well (from the Volokh Conspiracy point of view) for former Pres. Trump in court this morning.
Trump’s lawyer indicated he will not rely on the points advanced by the Calabresi brief (filed on Prof. Calabresi’s behalf by longtime Republican bigot Gene Schaerr).
The judges — including the Republican — are questioning whether “take care that laws are faithfully executed” means “above the law,” and asking Trump’s lawyer why Trump argued during impeachment that the matter should be handled in a criminal court.
Carry on, clingers . . .
Start:
AI Companies: We must do something.
AI Compaines: This is something!
AI Companies: We must do this!
EU Regulators: We must do something!
EU Regulators: This is something!
EU Regulatore: We must do this!
Goto Start:
By the way, in case people want to actually read the thing before having opinions, the final text of the proposal isn't available yet. On 9 December the Council and the Parliament reached a political agreement, and that agreement still has to be turned into official legal text in all official languages. But the press release about the political agreement is here: https://www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificial-intelligence-act-deal-on-comprehensive-rules-for-trustworthy-ai
...and the dossier on the European Parliament website is here, with the original proposal as drafted by the Commission, and the negotiating version adopted by the Parliament in June: https://oeil.secure.europarl.europa.eu/oeil/popups/ficheprocedure.do?reference=2021/0106(COD)&l=en
Sounds like a job for AI.
So let me get this straight. The people who are currently yammering about AI "safety" are referring to something like a Trump getting elected? They're just worried that AI won't be sufficiently biased and censored toward their favored politics?
Whatever gave you that idea?
The reasons for the proposal are right on page 1. Easy peasy, no searching necessary: https://www.europarl.europa.eu/RegData/docs_autres_institutions/commission_europeenne/com/2021/0206/COM_COM(2021)0206_EN.pdf
"It is in the Union [sic] interest to preserve the EU’s technological leadership"
Hilarious. They have no "leadership", just a tech backwater.
If you say so.
No, revenue and market caps say so.
Not a booming leader like Ohio?
Ohio doesn't seek to regulate the world though.
Neither does the EU. The EU regulates the EU. What other countries then do is up to them.
Against this political context, the Commission puts forward the proposed regulatory framework on Artificial Intelligence with the following specific objectives:
ensure that AI systems placed on the Union market and used are safe and respect existing law on fundamental rights and Union values;
ensure legal certainty to facilitate investment and innovation in AI;
enhance governance and effective enforcement of existing law on fundamental rights and safety requirements applicable to AI systems;
facilitate the development of a single market for lawful, safe and trustworthy AI applications and prevent market fragmentation."
Right. So just have the government control AI and the market for AI, and ban any AI that does or says things we don't like i.e. goes against our "values" or is owned by people we don't like.
I doubt the EU have a proper handle on the problem, but AI is just a fucking algorithm, it doesn't have rights, citizens have rights, and citizens would prefer if shitty algorithms didn't spread all over the internet and render it nigh unusable, which is one of the worst-case scenarios, not that 'AI will destroy humanity' bullshit the techbros are trying to snow regulators with.
That's not what people are talking about when they discuss AI safety.
It's a whole sub-discipline of research, actually, so saying AI safety has a goal is not even wrong.
Sure it is. They need to stop "mis/disinformation" as if this is some new problem rather than as old as human history, and as if they have a scintilla of credibility in what and how they are labeling misinformation.
Just a continuation of the hoopla from the last 8+ years, threatening and cajoling tech companies to increase their massive censorship and thought curation efforts. Google, YouTube, etc have all changed drastically.
When you rely so heavily on dis/misinformation, this must be a dismaying prospect.
LLM's aren't even really AIs by a lot of standards, but making sure they don't, say, post fake legal citations, is not a sinister plot.
It is the government's job to ensure fake legal citations don't get posted. See, .e.g., Your Mom v. Beached Whale, 499 F. 3d 1023 (9th Cir. 2008)
AI, including AI safety, is funded by federal grants. Grants are not contracts – they are government funds given towards the production of a public good, like scientific papers published for the general public/scientific community.
Meanwhile, the FTC is consulting on a possible regulation of AI-based voice cloning: https://twitter.com/linakhanFTC/status/1744765781924106566
AI is already incredibly expensive and requires a huge infrastructure to operate so who knows when and how it’ll burn out, but now they’re complaining that if they have to pay copyright holders for works they steal from it would be completely unviable. Which is just to say that in the face of a global warming crisis threatening the future of the planet we live on, tech ‘innovaters’ are doing their absolute and complete worst to ignore the problem and make things that are expensive and clever and crap. I question whether any of these supposed nerds read any science fiction at all.
Consider the question begged.
It was stealing when spotty teenagers were copying casette tapes, it's good business practice when wealthy corporations do it on industrial scales.
Fair use is an actual thing.
But this is taking the piss.
"AI is already incredibly expensive and requires a huge infrastructure to operate"
I'm running a latent diffusion model AI on a five-year-old midrange PC. The interface was written by one guy and it uses the LMU Munich model, both of which are free to the public. Admittedly, it takes a lot of technical know-how to set up, by which I mean following a step-by-step guide and doing all of the steps.
O/T
Trump lawyer tells appeals court president can murder political rivals without being prosecuted — unless impeached and convicted by Senate
"The stark question was posed to Trump’s attorney John Sauer by Judge Florence Pan: Was a president immune from prosecution for any unlawful act, at all? Could a president order his political rivals to be assassinated by Seal Team 6 as an official act? Could he sell pardons at his pleasure if he saw fit and then face no consequences for his actions?
'He would have to be impeached and convicted first,' Sauer replied . . . . "
https://lawandcrime.com/high-profile/trump-lawyer-tells-appeals-court-president-can-murder-political-rivals-without-being-prosecuted-unless-impeached-and-convicted-by-senate/
Any Trumpers want to try and defend that?
Sauer's argument was far more radical than this. Sauer's argument insanely was that the president would have to be impeached and convicted for that specific thing first.
If Trump were impeached and convicted for the events of J6, and then a month after he's removed from office we learn that he put out a hit on Joe Biden or accepted cash from Vladimir Putin in exchange for classified information, he would be absolutely immune from prosecution according to Sauer because he was impeached and convicted for something entirely different.
Welp, ball's in Biden's court now.
Aaron Rodgers may be talking himself into a defamation litigation (and judgment) on live television on ESPN as I write this.
He also may be forcing ESPN's hand with respect to his continued participation with the Pat McAfee show, which seems to be a downscale amateur hour produced by and for uninformed, antisocial imbeciles.
"Aaron Rodgers may be talking himself into a defamation litigation (and judgment) on live television on ESPN as I write this."
Wow, the EU's AI Act affects defamation litigation in the US? That's a pretty broad law.