-
The EU passed new rules banning certain AI applications and regulating others based on perceived risk level
-
Analysts said banned apps make up only a small portion of the market
-
In a middle-of-the-road scenario, IDC said spending on AI should only slow slightly compared to previous revenue forecasts
It was only a matter of time, really. The European Union (EU) officially put artificial intelligence (AI) on a leash this month, passing a new law called the EU AI Act which bans certain applications, regulates others based on perceived risk-level and implements transparency requirements for general purpose systems.
Analysts from Gartner and IDC told Silverlinings the law could lead to fines for organizations that violate the new rules but noted it’s not really expected to dramatically cool overall AI spending.
Banned applications include those which scrape facial images from the internet or CCTV to create facial recognition databases (looking at you, Clearview AI). Emotion recognition in schools and the workplace, predictive policing and systems that use sensitive characteristics for biometric categorization also got the boot.
AI systems used for critical infrastructure, education and vocational training, essential services such as healthcare and banking, and certain law enforcement applications are deemed high-risk under the law. Thus, these are subject to regulations mandating the use of logs, risk assessments and human oversight. They will also be required to accept complaints from consumers and provide explanations about decisions that were made based on high-risk AI systems.
All general purpose AI systems must now meet new transparency requirements, which call for them to comply with EU copyright law and provide visibility into the content used for model training.
Impact
IDC Research Director Ewa Zborowska told Silverlinings, “The AI Act targets providers placing AI systems in the EU markets, irrespective of whether those providers are based in the region or in a third country.”
Why does that matter?
Well, in a note to clients, Gartner Distinguished VP Analyst Avivah Litan and team pointed out the law includes a “progressive sanctioning scale” that ties fines to the severity of each violation. Gartner’s team added that since the AI Act is closely tied to the EU’s General Data Protection Regulation (GDPR), organizations that violate the former will also likely pay fines related to the latter.
“SMEs and start-ups will likely be fined in a way that is proportional to their size compared to larger technology firms,” they wrote. Fines can range up to 7% of the total worldwide annual revenue of the previous financial year, amounting to a hefty chunk of change.
That said, Zborowska noted technology vendors and those deploying high-risk systems are among those likely to be impacted by the regulations.
Litan and crew warned that the law could make it harder for European AI companies to compete with overseas vendors by slowing time to market and forcing them to spend more on legal services.
The act could “temporarily drive investment in AI to other regions, where regulations are less demanding,” Gartner's team said.
Gartner also speculated that the Act could slow European adoption of AI technologies, but longer term noted that it could help “harmonize third-party vendor risk assessment processes and controls across all organizations.”
Forecast tweaks
Zborowska said IDC has modeled three different scenarios for how the law might impact AI spending: Best-case, Balanced and Worst-case. She said the Balanced model calls for annual AI software revenue to fall between 5-6% behind where it would have been over the 5-year forecast period if the Act wasn’t passed.
In the Best-case scenario, revenue forecasts wouldn’t change at all, but in the Worst-case, revenue would plunge 19-20% below the forecast levels. Public service spending on AI software is likely to be the hardest hit, she said.
Zborowska noted banned and high-risk use cases “only make up a small proportion of the overall AI market today.”
She added, “It is worth considering that the impact on AI software market will not happen in a vacuum. Whatever the impact on AI software market, it will also be reflected in AI-related services spending and hardware.”