Prepare your audits: EU Commission approves first-of-its-kind AI Act
(2024/05/21)
- Reference: 1716303551
- News link: https://www.theregister.co.uk/2024/05/21/prepare_your_audits_eu_commission/
- Source link:
The EU Council has given final approval to the bloc's landmark AI Act, setting the stage for enactment of a benchmark first-of-its-kind AI law across Europe.
The EC [1]describes the AI Act as taking a "risk-based approach" to AI regulation, meaning that the greater the risk an AI product could harm society, the more regulations it'll have to follow in order to do business in the EU. At the same time, the EC said it hopes the AI Act will open the door to investment in the right kinds of AI – safe, transparent ones.
For that reason, "high-risk" AI systems (those that are used in critical sectors, as well as in law enforcement, for identification and other purposes) will have to submit to regular audits, pass a fundamental rights impact assessment, and register in a central database of high-risk AI systems.
[2]
"With the AI act, Europe emphasizes the importance of trust, transparency and accountability when dealing with new technologies while at the same time ensuring this fast-changing technology can flourish and boost European innovation," said Belgian secretary of state for digitization, Mathieu Michel.
[3]
[4]
Along with classifying and regulating various types of AI and the companies that make them, the AI Act also declares several types of AI verboten in the bloc.
The UK wants you to acknowledge its AI victory, too!
Talk about being overshadowed: The UK has an AI victory to celebrate as the country just got 16 AI firms with global reach to commit to developing safe AI models, and the EU goes and passes the AI Act.
While Europe touts its world-first law, The UK and the Republic of Korea said they've managed to get Amazon, Google, IBM, Microsoft, Meta and others to agree to the Frontier AI Safety [5]Commitments . The agreement would require AI companies to properly assess the risks inherent in their latest models and to not publish them if the risk of misuse is too great.
Much of the document builds on the [6]Bletchley Declaration signed last year, the UK government said.
"These commitments ensure the world's leading AI companies will provide transparency and accountability on their plans to develop safe AI," said UK prime minister Rishi Sunak. The agreement doesn't appear to be in any way binding.
"Cognitive behavioural manipulation and social scoring will be banned from the EU because their risk is deemed unacceptable," the Commission said. "The law also prohibits the use of AI for predictive policing based on profiling and systems that use biometric data to categorise people according to specific categories such as race, religion, or sexual orientation."
Future potential AGIs are included in the rule as well, and the act provides exemptions for systems used for military/defense and research purposes.
[7]EU lassos tech giants in bid to rein in the AI Wild West
[8]Republican senators try to outlaw rules that restrict Wall Street's use of AI
[9]Tech world won't have long to fall in line when EU signs off on AI Act
[10]Law prof predicts generative AI will die at the hands of watchdogs
The Act also calls for several new offices to be set up to handle administration of the new law. AI Offices at the Commission and parliamentary level will be created, as will a panel of experts and a stakeholders advisory forum. Fines for noncompliance are in the mix as well, and will be "set as a percentage" of global annual turnover or a predetermined amount, whichever is greater.
Passed by the European Parliament in [11]December , the AI Act hasn't been without its critics, who've said it appeared written with Big Tech in mind. That said, it's on the books now, will go into force in two years, and is likely to have quite a reach.
"Companies outside the EU who use EU customer data in their AI platforms will need to comply," lawyer Patrick van Eecke [12]told Reuters of the passing of the Act. "Other countries and regions are likely to use the AI Act as a blueprint, just as they did with the GDPR." ®
Get our [13]Tech Resources
[1] https://www.consilium.europa.eu/en/press/press-releases/2024/05/21/artificial-intelligence-ai-act-council-gives-final-green-light-to-the-first-worldwide-rules-on-ai/?utm_source=brevo&utm_campaign=AUTOMATED%20-%20Alert%20-%20Newsletter&utm_medium=email&utm_id=320
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.gov.uk/government/publications/frontier-ai-safety-commitments-ai-seoul-summit-2024/frontier-ai-safety-commitments-ai-seoul-summit-2024
[6] https://www.theregister.com/2023/11/01/uk_ai_summit/
[7] https://www.theregister.com/2024/01/03/eu_ai_legistlation/
[8] https://www.theregister.com/2024/02/07/republican_sec_ai_finance/
[9] https://www.theregister.com/2024/01/24/eu_ai_act_timetable/
[10] https://www.theregister.com/2024/04/24/generative_ai_death/
[11] https://www.theregister.com/2023/12/11/eu_agrees_ai_act/
[12] https://www.reuters.com/world/europe/eu-countries-back-landmark-artificial-intelligence-rules-2024-05-21/
[13] https://whitepapers.theregister.com/
The EC [1]describes the AI Act as taking a "risk-based approach" to AI regulation, meaning that the greater the risk an AI product could harm society, the more regulations it'll have to follow in order to do business in the EU. At the same time, the EC said it hopes the AI Act will open the door to investment in the right kinds of AI – safe, transparent ones.
For that reason, "high-risk" AI systems (those that are used in critical sectors, as well as in law enforcement, for identification and other purposes) will have to submit to regular audits, pass a fundamental rights impact assessment, and register in a central database of high-risk AI systems.
[2]
"With the AI act, Europe emphasizes the importance of trust, transparency and accountability when dealing with new technologies while at the same time ensuring this fast-changing technology can flourish and boost European innovation," said Belgian secretary of state for digitization, Mathieu Michel.
[3]
[4]
Along with classifying and regulating various types of AI and the companies that make them, the AI Act also declares several types of AI verboten in the bloc.
The UK wants you to acknowledge its AI victory, too!
Talk about being overshadowed: The UK has an AI victory to celebrate as the country just got 16 AI firms with global reach to commit to developing safe AI models, and the EU goes and passes the AI Act.
While Europe touts its world-first law, The UK and the Republic of Korea said they've managed to get Amazon, Google, IBM, Microsoft, Meta and others to agree to the Frontier AI Safety [5]Commitments . The agreement would require AI companies to properly assess the risks inherent in their latest models and to not publish them if the risk of misuse is too great.
Much of the document builds on the [6]Bletchley Declaration signed last year, the UK government said.
"These commitments ensure the world's leading AI companies will provide transparency and accountability on their plans to develop safe AI," said UK prime minister Rishi Sunak. The agreement doesn't appear to be in any way binding.
"Cognitive behavioural manipulation and social scoring will be banned from the EU because their risk is deemed unacceptable," the Commission said. "The law also prohibits the use of AI for predictive policing based on profiling and systems that use biometric data to categorise people according to specific categories such as race, religion, or sexual orientation."
Future potential AGIs are included in the rule as well, and the act provides exemptions for systems used for military/defense and research purposes.
[7]EU lassos tech giants in bid to rein in the AI Wild West
[8]Republican senators try to outlaw rules that restrict Wall Street's use of AI
[9]Tech world won't have long to fall in line when EU signs off on AI Act
[10]Law prof predicts generative AI will die at the hands of watchdogs
The Act also calls for several new offices to be set up to handle administration of the new law. AI Offices at the Commission and parliamentary level will be created, as will a panel of experts and a stakeholders advisory forum. Fines for noncompliance are in the mix as well, and will be "set as a percentage" of global annual turnover or a predetermined amount, whichever is greater.
Passed by the European Parliament in [11]December , the AI Act hasn't been without its critics, who've said it appeared written with Big Tech in mind. That said, it's on the books now, will go into force in two years, and is likely to have quite a reach.
"Companies outside the EU who use EU customer data in their AI platforms will need to comply," lawyer Patrick van Eecke [12]told Reuters of the passing of the Act. "Other countries and regions are likely to use the AI Act as a blueprint, just as they did with the GDPR." ®
Get our [13]Tech Resources
[1] https://www.consilium.europa.eu/en/press/press-releases/2024/05/21/artificial-intelligence-ai-act-council-gives-final-green-light-to-the-first-worldwide-rules-on-ai/?utm_source=brevo&utm_campaign=AUTOMATED%20-%20Alert%20-%20Newsletter&utm_medium=email&utm_id=320
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZkzFJo6HzKoFdNEa13EYhQAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.gov.uk/government/publications/frontier-ai-safety-commitments-ai-seoul-summit-2024/frontier-ai-safety-commitments-ai-seoul-summit-2024
[6] https://www.theregister.com/2023/11/01/uk_ai_summit/
[7] https://www.theregister.com/2024/01/03/eu_ai_legistlation/
[8] https://www.theregister.com/2024/02/07/republican_sec_ai_finance/
[9] https://www.theregister.com/2024/01/24/eu_ai_act_timetable/
[10] https://www.theregister.com/2024/04/24/generative_ai_death/
[11] https://www.theregister.com/2023/12/11/eu_agrees_ai_act/
[12] https://www.reuters.com/world/europe/eu-countries-back-landmark-artificial-intelligence-rules-2024-05-21/
[13] https://whitepapers.theregister.com/
Turing Test
Conundrum1885
The risk here is that first you have to prove something is an AI.
What happens when a physics paper or some other publication is rejected because the reviewer(s) believe that it is
the work of an artificial intelligence when in fact it is not, and science then gets set back?
Or worse, a human researcher is wrongly accused of plagiarism, has their grant or other finance taken away and it later
emerges that they were innocent and the accusation was itself made by an AI stringing together the facts in a way that
logically makes sense but is still incorrect.
Re: Turing Test
Doctor Syntax
"
The risk here is that first you have to prove something is an AI."
Take the vendor's word for it. Having made a big fuss about their wonderful AI they'll be in no place to claim the legislation doesn't apply. Conversely the need to make the big fuss will ensure they do just that.
I've just skimmed the act, but it seems to me like the tough bits only apply to AI systems that are used to make important decisions about people (law enforcement, employee evaluation, surveillance, etc), or that control safety-critical systems. I think there's something about big commercial providers. But free and open-source systems that are not high-risk are explicitly exempt from the whole regulation. That's actually more reasonable than I feared.