EU AI Act
What the AI Act EU has already altered for companies
Among the first artificial intelligence legislation to be implemented in one of the biggest marketplaces in the world is the AI Act EU.
What changes can businesses expect?
Describe the EU AI Act
The first significant market to establish new AI regulations is the European Union (EU). As stated by EU authorities, “the goal is to establish the EU as a worldwide Centre for reliable AI.”
What is the EU AI Act
Using a risk-based methodology, the AI Act groups applications depending on how likely they are to endanger people’s safety and basic rights. Rules for general-purpose AI (GPAI) models, criteria for developing and implementing particular high-risk AI systems, and a ban on specific AI practices that are judged to pose unacceptable risk are some of the most significant clauses.
Known as the “minimal risk” category, AI systems that do not fit into any of the risk categories outlined in the AI Act EU are exempt from its restrictions, however some may still need to adhere to other applicable laws and transparency standards.
According to Bruno Massot, Vice President, Assistant General Counsel, and Head of Legal at IBM Europe, “it is a form of regulatory pragmatism that aims to adapt constraints with the level of risk.”
Certain AI applications that pose a threat to people’ rights will be prohibited under these new regulations. Predictive policing, emotion recognition in the workplace and classroom, untargeted facial picture scraping from the internet or CCTV footage to build facial recognition databases, and biometric categorization systems based on sensitive features are some examples.
European Union AI Act
On April 22, the European Parliament passed the AI Act EU, and on May 21, the EU Member States did the same. The law went into effect on August 1st, after several clauses were gradually implemented. It was first published in the European Union’s Official Journal on July 12th.
EU Act On AI
The timeline for the EU AI Act will move extremely quickly
Everything is not surprising, even though everything is happening quite quickly. It’s long been known about the drafts. Given how swiftly technology is developing, it’s also critical to move quickly, according to Massot.
Sixth month: Prohibitions against AI techniques go into effect. Codes of practice come into effect in month nine. General-purpose AI regulations, including governance, take effect in month twelve. The high-risk AI system regulations will go into force after 24 months. In conclusion, the regulations pertaining to AI systems that are either products or safety components of products subject to special EU legislation will come into effect 36 months following the laws’ introduction.
The EU AI Act
Which actions ought firms to take?
Dasha Simons, Managing Consultant, Trustworthy AI at IBM, notes that “there is a sense of urgency.” “Somethings taking two or three years to prepare could appear excessive. However, it’s not that lengthy if you’re a global or multinational company with numerous AI systems located all over the world.
Where should they begin then? According to Simons, the first stage is to decide which regulations will be relevant to your company.
The AI Act EU lays out distinct guidelines and terminology for importers, providers, and delayers. You will have to adhere to distinct regulations based on your company’s function, according to Simons. Are you purchasing models and merely giving them a new name without altering the model itself? Or are you really adjusting the model quite a little bit? Essentially, this is the initial phase.
- The second stage would be to ascertain which AI systems are in operation at your company and the degrees of risk attached to each one.
- “Knowing where you want to focus at first is important,” says Simons.” Many businesses have no idea what kind of AI models and systems they have in development or in production.”
- Businesses can identify their priorities by completing this assessment, which evaluates the high-risk systems after the prohibited ones.
- “This will assist in identifying the procedures that must be implemented when new AI systems are introduced, and ensure that compliance with the AI Act EU is done proactively rather than reactively.”
- According to Simons, the rules specify that companies must use AI strategically and that the C-suite must actively participate in this discussion.
You require technical know-how; in addition, it can be helpful to have a Chief Privacy Officer with experience implementing GDPR. While it’s important to have that knowledge available, Simons also believes that the C-level should establish accountability and strategy.
What effects will the EU AI Act have on companies?
The EU aims to set a global AI development standard. The GDPR of 2018 gave the EU a comprehensive data privacy and security law.
Parliamentarian Dragos Tudor ache, the AI Act’s main negotiator, called it “the first regulation in the world that is laying a clear roadmap on a safe and human-centric development” at a news conference.
Regarding the use of AI, numerous policies are being adopted globally
“Because it would be difficult [for businesses] to develop different systems with diverging restraints, consistency is important,” adds Massot.
An interesting development will be the requirement for organizations and enterprises operating in the EU to teach their users or workers about AI, according to Hans-Petter Dalen, IBM’s EMEA Business Leader for Watson and embeddable AI.
It’s interesting to note that there is a section on AI literacy in the AI Act EU. Hence, in order to use AI in the EU single market, any business or organization must provide a specific degree of education to both its users and workers. According to Dalen, “raising the fundamental understanding of what AI is is a very desirable thing. They don’t know what the level is yet.
“HR systems, which can currently screen resumes and suggest applicants, are one example in that regard. There are seven crucial requirements that must be met in order to proceed with that high-risk use case. Are you knowledgeable enough about artificial intelligence (AI) as the person purchasing that program to know what questions to ask about the algorithms? asks Dalen.
EU AI Act High Risk
The technological criteria that the AI Act will mandate remain mostly undetermined
The regulation itself lays out seven very loosely worded standards that must be met in order to comply with high-risk use scenarios. Ten requests for technical standards have been generated by those seven requirements, and two European standard decision organizations are currently working on developing those requests, says Dalen.
They anticipate that putting the technical standards into practice will be the quickest and least expensive way to achieve conformity, as they will provide us with clarity on the actual requirements from the regulation.”
