I doubt it’s the end of AI access in Europe. Even if they do fine them 3% of global turnover, that’s probably a small enough fraction of their European revenue that it’s worth the risk. After all if they drop European traffic it’s not like it’ll change their fixed costs, and the marginal cost to serve Europe is probably relatively small compared to their fixed costs.
I do think it’s most likely that they will end up fining the AI companies quite a bit under this act. It seems like a bit of a pattern with European legislation that disproportionately impacts American tech companies.
Perhaps the fines themselves act as a de facto AI tax that some people have been trying to promote in the political sphere, and we'll see how successful it is.
It's a (successful?) lobbying attempt by Mistral. They are falling behind on tech and they want to win on another stupid EU law. You know what will happen next? There will be an exception carved in for "developing countries" ... like China.
I really oppose the notion that “European legislation … disproportionately impacts American tech companies”. US tech has a profound impact on the lives of everyday Europeans while the US has almost no checks and balances in place for the tech industry. Europeans are ought to be able decide how they want to live their lives and build a society. This must include deciding how much influence American tech companies shall be allowed to have by passing laws for Europe.
Roughly half of American states have passed extensive internet regulation in the last few years to the point of entire businesses exciting their states, and that's the conservative parts of America.
The American public, of various creeds and political orientations is as, if honestly not more skeptical of US tech companies. They simply have no control over their government, which is captured by tech CEOs.
The recent anti datacenter activism extends to IIRC over 40 states now and has roughly 70% largely bipartisan support.
The whole premise ignores how much the same companies are fined and settling for in the US, for many of the same reasons: privacy abuse, market abuse, societal abuse, worker abuse, etc. Pound for pound the recent Meta settlement has certainly skewed the numbers much higher for the US too.
Isn't part of " Europeans are ought to be able to decide how they want to live their lives" that they have the choice of using products of US tech companies?
Claude slop article aside, this act is one reason why some people say Europe is an open air museum, as it'd rather regulate itself to obscurity while its great works are behind it. Where is Mistral, Le Chaton Fat? They're not competitive anymore. Meanwhile China reaches ever forward because they know that speed and endurance is everything in a race, not shooting oneself in the foot.
The key problem for Europe is that they don't want to take strong enough actions against US tech to let their own actors develop. Which is how China has reached tech dominance, and how even a shithole country like Russia can have their own “tech” industry.
Even in a hypothetical world where they somehow killed off US tech completely, they still wouldn't have their own actors develop. There is no collective will, no risk taking mindset. The VCs are cowardly with capital, and people don't follow through on bright ideas. Taxes and regulations suffocate any progress and no one wants to do anything about it.
The EU would rather cruise off into perceived comfort and safety rather than attempt to stay relevant economically and on the world stage. Only when the region is unable to afford its own welfare will it start wondering where things went wrong.
The fact that Russia, with its notably worse institutional framework for business has more relevant tech companies than the EU speeks volume about the value of your argument.
It doesn't matter how good or bad your legislation is when you are always facing an uphill battle against a better capitalized company serving a bigger market, especially in a sector which, despite its name, has very little technology in it, and hence little barriers besides marketing and network effect.
The idea was that technology is now powerful enough, everything need not be just competition any more. The world could also work to let every human have a good life. Neither China nor America truly seem to have that goal, despite it being both possible and ethical. Europe at least did, based on our experiences of the horrors of the early/mid C20.
It's true that in the last 10 years that vision feels like it has crumbled. And the inevitableness of competition is coming back. International law and human rights are dissolving - rule based on power. This works well for anyone while it works for them - except for our souls, and longer term... It's hubristic and eventually empires fall. Europe has already experienced this, and remembers it.
Whether these empires will fall to global economic collapse from lack of co-operation, or environmental destruction from lack of charging for externalities, or from global war, or simply from out of control AI, remains to be seen.
But never question, it is possible to avoid all those. But we have to try to, not just cheaply go "competition good", because of the short term thrill of the win. Competition is good, and so is cooperation.
This is a very naive viewpoint that does not understand that superintelligence is a race between the West and China.
While the EU is twiddling its thumbs about copyright, it is shooting its own side in the foot during the most important race in human history.
The pure absurdity of this is a reflection of how the EU routinely chooses short term safety and good vibes over long term prosperity and success. This mindset is why the EU has been utterly unable to compete for years -- and now, they want to impose their failed model on the actually successful Western companies.
China and America's goals will not be influenced one bit by the EU bureaucracy. The only effect it will have it will make EU fall behind even further.
We'd need quite some changes in the EU to focus on actually winning at something again. Winning at the expense of some of the lofty ideals. I'd rather EU have a decent frontier model lab that plays to win, and steps on toes while doing so, than have a very elaborate bureaucracy that tries to prevent that from happening. Its no surprise the bureaucrats have the exactly opposite goals, and its sad that most of the population won't see the consequences of this conflict of interest until it's too late to do anything -- because the Americans and the Chinese have near AGI capabilities, plentiful energy sources, and, really, not a lot of reasons to remember where Europe is on the map.
> I'd rather EU have a decent frontier model lab that plays to win, and steps on toes while doing so, than have a very elaborate bureaucracy that tries to prevent that from happening
What in the AI Act makes it hard to develop frontier models?
Its just a tax and an annoyance, and in a very competitive and difficult economy.
I am already receiving notices from some EU countries that they will want sit downs on how we use AI, and apparently they have the right to demand access to detailed designs and source codes. Fine, whatever, a waste of time, but why would anyone start an AI company here, knowing that there will be bureaucrats focused on making that startup's life harder? Considering AI talent is mobile I'd think a lot of such talent will just hop on a plane and have their startup somewhere less overregulated.
GDPR would be enough if money was the aim. Enforcement is pretty thin with that as well, not as if there weren't enough companies out there that can and need to be fined.
The article only lists US labs, I wonder if that was an oversight. Chinese models are also available to Europeans through API providers. Surely they risk fines too.. right?
The regulations apply foremost to closed models. Chinese companies mostly release "open models".
And regarding fines, well the track record for EU fines against any big tech companies, US or China, is abysmal. Even if the companies are fined, the fine goes uncollected or is slashed after the big press release.
Why would they not apply to all models? The models with downloadable weights aren’t “open”. Meaning no one can actually give a summary about their training content and security. Which is what they’re asking US labs.
"The obligations set out in paragraph 1, points (a) and (b), shall not apply to providers of AI models that are released under a free and open-source licence that allows for the access, usage, modification, and distribution of the model, and whose parameters, including the weights, the information on the model architecture, and the information on model usage, are made publicly available. 2This exception shall not apply to general-purpose AI models with systemic risks."
I absolutely hate those bureaucrats, people who never built anything and just want to control, review, and develop their cottage industries of well paid lawyers one needs to retain to operate in the market. Same as with GDPR - absolutely no positive impact on actual privacy, but a lot of burden on everyone.
The worst part is, those people can't be voted out. Because of how EU works, the bureaucrats have an upper hand over the elected officials, and because of lavish compensation for the said elected officials, mostly bad people get elected anyway - those whos domestic careers have ended, or those whose parties want out of active politics, and somewhere "safe" where they can earn good money without causing any trouble. Thats a horrible system.
> Same as with GDPR - absolutely no positive impact on actual privacy, but a lot of burden on everyone.
Confidently wrong on this one; it has been very helpful for me to both extract data as well as exercise my right to be forgotten from various providers that would otherwise have no obligation to delete my data.
The counter argument I'd make is: the current state of surveillance in the EU and elsewhere is companies try to stay on the thin edge between plausibly-aggregate-data-but-not-really -- and PII -- and skirt gdpr.
This sort of data is not subject to data rights requests, and is quite ok for most things - the insurance company knows what kind of hobbies does someone quite like you but plausibly not only you have, the pr0n companies know what do you typically watch, and they are free to exchange those semi-aggregates with each other if it makes sense to them. This requires no cookies, no consent, and is kept precisely on the right side of plausibly-non-identifying-aggregates. There is a substantial industry of companies brokering such data.
That is my argument against considering gdpr a good thing. People think it protects their privacy, but in reality it just polices the good actors, and does literally nothing to thwart sophisticated malicious actors. One of the reasons is, the people who legislated gdpr do not understand science nor tech -- nor business.
p.s. i'd add, as someone who runs tech for a company with about 700k daily active users, people who exercise data subject rights are the extreme minority. A handful of cases per month. So you have a legislation that largely misses its stated mark, imposes heavy bureaucratic burden on everyone, and serves a small minority. I am not against the concept of data privacy, I just say that the current setup is ineffectual.
Started reading, saw “The timing is the story”, and my LLM-generated-spider-sense tingled. The piece is informative, but I now have no idea if the arguments are human or AI driven
One can and must discuss and challenge all these regulations, but this seems better than just sliding carelessly into some sort of mega-corp dystopia. It is good to have some balance of power. Is the EU perfect? Absolutely not, but neither are the corporates. Ideally, compromises will be found (and mostly they are found). Honestly, what’s the worst that can happen here? If this slows down the European economy, that’s a trade-off Europe has chosen to take (and yes, I’m very aware of the difficulties of fair representation in such a complex entity as the EU). We now have three big power blocks with very different politics and governance models. Isn’t that so much better than just playing one massive bet on a few big AI players? Honestly, I think it is exciting!
> Honestly, what’s the worst that can happen here?
Regulation that cannot practically be implemented or complied with (example: forks of open models), this leads to selective enforcement and by extension corruption and loss of rule of law.
Regulation before the market has even settled down, we don't know if AI will be datacenter only, or open models everybody forks, modifies and runs locally.
> Regulation before the market has even settled down
The counter point would be: the major AI providers have seemingly also lost control of their creation. So at the very least both sides are doing things before they fully understand what they are doing. Also, looking at the (planned) IPOs and the share of AI in S&P 500, one could argue that while the market hasn’t settled, it already is a massive market. When should we start regulating? Eventually it will be too late
> Regulation that cannot practically be implemented or complied with
Yes. Having worked with one EU authority, it’s still people who eventually do the enforcing. That has the risk of corruption but I have also seen repeatedly that they won’t do something if it violates common sense. As always, there is room for interpretation and enforcement and it will be used. Like with any good governance. And companies still can (and do!) sue
53 comments
[ 0.20 ms ] story [ 3.3 ms ] threadI do think it’s most likely that they will end up fining the AI companies quite a bit under this act. It seems like a bit of a pattern with European legislation that disproportionately impacts American tech companies.
The American public, of various creeds and political orientations is as, if honestly not more skeptical of US tech companies. They simply have no control over their government, which is captured by tech CEOs.
The recent anti datacenter activism extends to IIRC over 40 states now and has roughly 70% largely bipartisan support.
But that only explains that Americans want regulation, not that they want EU regulation.
Americans put innovation and result ahead while Europe prefers precaution/fairness first.
There are real costs for Europe in terms of productivity because of this. Take a look at the Draghi report.
Also see European founders relocating and citing regulation as a problem when it comes to access to capital.
The EU would rather cruise off into perceived comfort and safety rather than attempt to stay relevant economically and on the world stage. Only when the region is unable to afford its own welfare will it start wondering where things went wrong.
We could rephrase this as "the shareholders aren't maximizing their value and the average Joe doesn't have a problem with it".
It doesn't matter how good or bad your legislation is when you are always facing an uphill battle against a better capitalized company serving a bigger market, especially in a sector which, despite its name, has very little technology in it, and hence little barriers besides marketing and network effect.
It's true that in the last 10 years that vision feels like it has crumbled. And the inevitableness of competition is coming back. International law and human rights are dissolving - rule based on power. This works well for anyone while it works for them - except for our souls, and longer term... It's hubristic and eventually empires fall. Europe has already experienced this, and remembers it.
Whether these empires will fall to global economic collapse from lack of co-operation, or environmental destruction from lack of charging for externalities, or from global war, or simply from out of control AI, remains to be seen.
But never question, it is possible to avoid all those. But we have to try to, not just cheaply go "competition good", because of the short term thrill of the win. Competition is good, and so is cooperation.
While the EU is twiddling its thumbs about copyright, it is shooting its own side in the foot during the most important race in human history.
The pure absurdity of this is a reflection of how the EU routinely chooses short term safety and good vibes over long term prosperity and success. This mindset is why the EU has been utterly unable to compete for years -- and now, they want to impose their failed model on the actually successful Western companies.
We'd need quite some changes in the EU to focus on actually winning at something again. Winning at the expense of some of the lofty ideals. I'd rather EU have a decent frontier model lab that plays to win, and steps on toes while doing so, than have a very elaborate bureaucracy that tries to prevent that from happening. Its no surprise the bureaucrats have the exactly opposite goals, and its sad that most of the population won't see the consequences of this conflict of interest until it's too late to do anything -- because the Americans and the Chinese have near AGI capabilities, plentiful energy sources, and, really, not a lot of reasons to remember where Europe is on the map.
What in the AI Act makes it hard to develop frontier models?
I am already receiving notices from some EU countries that they will want sit downs on how we use AI, and apparently they have the right to demand access to detailed designs and source codes. Fine, whatever, a waste of time, but why would anyone start an AI company here, knowing that there will be bureaucrats focused on making that startup's life harder? Considering AI talent is mobile I'd think a lot of such talent will just hop on a plane and have their startup somewhere less overregulated.
And regarding fines, well the track record for EU fines against any big tech companies, US or China, is abysmal. Even if the companies are fined, the fine goes uncollected or is slashed after the big press release.
I am looking as the DSA here...
"The obligations set out in paragraph 1, points (a) and (b), shall not apply to providers of AI models that are released under a free and open-source licence that allows for the access, usage, modification, and distribution of the model, and whose parameters, including the weights, the information on the model architecture, and the information on model usage, are made publicly available. 2This exception shall not apply to general-purpose AI models with systemic risks."
https://ai-act-law.eu/article/53/
The why is under:
https://ai-act-law.eu/recital/102/
So the chinese models aren't exactly open like in the EU wants them to be
The worst part is, those people can't be voted out. Because of how EU works, the bureaucrats have an upper hand over the elected officials, and because of lavish compensation for the said elected officials, mostly bad people get elected anyway - those whos domestic careers have ended, or those whose parties want out of active politics, and somewhere "safe" where they can earn good money without causing any trouble. Thats a horrible system.
Confidently wrong on this one; it has been very helpful for me to both extract data as well as exercise my right to be forgotten from various providers that would otherwise have no obligation to delete my data.
This sort of data is not subject to data rights requests, and is quite ok for most things - the insurance company knows what kind of hobbies does someone quite like you but plausibly not only you have, the pr0n companies know what do you typically watch, and they are free to exchange those semi-aggregates with each other if it makes sense to them. This requires no cookies, no consent, and is kept precisely on the right side of plausibly-non-identifying-aggregates. There is a substantial industry of companies brokering such data.
That is my argument against considering gdpr a good thing. People think it protects their privacy, but in reality it just polices the good actors, and does literally nothing to thwart sophisticated malicious actors. One of the reasons is, the people who legislated gdpr do not understand science nor tech -- nor business.
p.s. i'd add, as someone who runs tech for a company with about 700k daily active users, people who exercise data subject rights are the extreme minority. A handful of cases per month. So you have a legislation that largely misses its stated mark, imposes heavy bureaucratic burden on everyone, and serves a small minority. I am not against the concept of data privacy, I just say that the current setup is ineffectual.
The age of written expertise influence is over.
Regulation that cannot practically be implemented or complied with (example: forks of open models), this leads to selective enforcement and by extension corruption and loss of rule of law.
Regulation before the market has even settled down, we don't know if AI will be datacenter only, or open models everybody forks, modifies and runs locally.
And we won't know for a couple for years to come.
> Regulation before the market has even settled down
The counter point would be: the major AI providers have seemingly also lost control of their creation. So at the very least both sides are doing things before they fully understand what they are doing. Also, looking at the (planned) IPOs and the share of AI in S&P 500, one could argue that while the market hasn’t settled, it already is a massive market. When should we start regulating? Eventually it will be too late
> Regulation that cannot practically be implemented or complied with
Yes. Having worked with one EU authority, it’s still people who eventually do the enforcing. That has the risk of corruption but I have also seen repeatedly that they won’t do something if it violates common sense. As always, there is room for interpretation and enforcement and it will be used. Like with any good governance. And companies still can (and do!) sue