The EU AI Act - the one about startups

The European Union's proposed Artificial Intelligence Act (AI Act) is set to introduce a comprehensive regulatory framework for AI systems, aiming to promote trustworthy AI and protect fundamental rights. While the intentions behind the AI Act are commendable, it is crucial to examine its potential impact on startups and venture capital, as these players are vital to the growth and innovation of the AI ecosystem.

That said, there is also a strong argument that slowing down this ecosystem may be beneficial to the greater good. Many of these points are covered elsewhere in the series, however the impact directly on startups & venture must be noted.

The Specifics

Article 5 - Prohibited AI practices:

The prohibition of certain AI practices, such as using AI for social scoring or exploiting vulnerabilities of specific groups, could limit the scope of AI applications startups can develop. This may be particularly challenging for startups working on innovative use cases that may fall under these prohibitions.

A startup developing a social media analytics tool that inadvertently exploits the vulnerabilities of elderly users by oversimplifying its user interface.

Article 6 and Annex III Classification of high-risk AI systems:

The broad range of AI systems classified as high-risk, including those used in critical infrastructure, education, employment, and law enforcement, could subject many AI startups to stringent requirements. Startups may struggle to identify whether their AI systems fall under the high-risk category, leading to uncertainty and potential compliance issues.

An AI-powered edutech startup that provides personalised learning recommendations, unknowingly falling under the high-risk category for educational AI systems.

Articles 16-23 - Obligations for providers/operators of high-risk AI systems:

The extensive obligations imposed on providers of high-risk AI systems, such as quality management, data governance, technical documentation, and conformity assessments, could be burdensome for startups with limited resources. Complying with these requirements may divert resources from product development and innovation.

A small startup creating an AI-based tool for optimising energy consumption in buildings, struggling to meet the extensive quality management and documentation requirements.

Articles 53-54 - Transparency obligations for general purpose AI models:

As mentioned in previous article. Startups developing or using general purpose AI models will need to comply with transparency obligations, such as providing model documentation and ensuring copyright compliance. These requirements could be challenging for startups that rely heavily on open-source AI models or lack the resources to produce extensive documentation.

A startup leveraging open-source general purpose AI models for sentiment analysis, facing difficulties in ensuring full copyright compliance and providing detailed model documentation.

Articles 60-61 - Testing of AI systems in real-world conditions:

While the AI Act allows for real-world testing of high-risk AI systems, the conditions and safeguards required, such as obtaining informed consent and providing oversight, may be difficult for startups to implement. This could limit the ability of startups to validate their AI systems in real-world scenarios.

A healthcare startup aiming to validate its AI-powered diagnostic tool in real-world clinical settings, encountering obstacles in obtaining informed consent from patients and ensuring adequate oversight.

Articles 85-87 - Remedies and reporting of infringements:

The AI Act grants affected persons the right to lodge complaints and provides protection for whistleblowers reporting infringements. Startups may face challenges in handling complaints and potential legal action, which could be costly and time-consuming.

A startup offering an AI-driven recruitment platform, facing a complaint from a job seeker alleging discriminatory practices, leading to a costly and time-consuming investigation.

The effects in general in the EU

1. Compliance costs: The AI Act introduces various requirements for high-risk AI systems, such as quality management, data governance, technical documentation, and conformity assessments. Adhering to these requirements may increase compliance costs for EU-based AI startups, particularly those developing high-risk AI applications.
2. Innovation and competitiveness: While the AI Act aims to promote trustworthy AI and protect fundamental rights, some startups may find the regulatory burden challenging, potentially slowing down innovation and reducing their competitiveness compared to startups in less regulated markets.
3. Legal certainty: The AI Act provides a clear legal framework for AI development and deployment in the EU, which could provide legal certainty for startups and encourage investment in the AI sector.
4. Support measures: The AI Act includes provisions to support SMEs and startups, such as regulatory sandboxes, reduced fees for conformity assessments, and prioritised access to AI regulatory sandboxes. These measures could help mitigate some of the compliance burdens for startups - in the EU!

Impact on venture capital:

1. Increased due diligence: Venture capital firms investing in EU-based AI startups may need to perform additional due diligence to ensure their portfolio companies comply with the AI Act, potentially increasing the time and cost of investments.
2. Sector-specific impact: The AI Act's risk-based approach may lead to venture capital firms focusing on AI startups in sectors with lower regulatory burdens, such as those not classified as high-risk AI systems.
3. Attractiveness of EU market: The AI Act could make the EU market more attractive for venture capital firms seeking to invest in trustworthy and ethically-aligned AI startups, as the regulation sets high standards for AI development and deployment. Equally, due to point 1. This could make the EU to risk laden and problematic leading to a balkanisation of services.

Impact on startups and venture capital in the US:

1. Competitive advantage: US-based AI startups may have a competitive advantage over their EU counterparts in the short term, as they are not subject to the same regulatory requirements. This could lead to faster innovation and lower compliance costs for US startups. However…
2. Market access: US-based AI startups that want to enter the EU market will need to comply with the AI Act, which may require additional resources and effort.
3. Regulatory spillover: The AI Act may influence the development of AI regulations in the US, as policymakers and industry stakeholders observe the impact of the EU's approach. This could lead to similar regulations being introduced in the US in the future. We’ve seen this before with the GDPR/CDDP. There is an argument, regardless of any vagaries in the law that the EU has done much of ‘the work’ and are first to market with these regulations so it would be sensible to align. Yeah right.

Impact on startups in the UK:

1. Divergence from EU regulations: As the UK has left the EU, it is not directly subject to the AI Act. However, the UK government has stated its intention to create its own AI governance framework, which may diverge from the EU's approach. This would be insane because...
2. Market access: UK-based AI startups that want to enter the EU market will need to comply with the AI Act, similar to US-based startups.
3. Attractiveness for investment: The UK's approach to AI regulation could influence its attractiveness for venture capital investment in AI startups, depending on whether it adopts a more or less stringent approach compared to the EU.

The EU AI Act is likely to have a significant impact on AI startups and venture capital in the EU, US, UK and beyond. While the regulation aims to promote trustworthy AI and protect fundamental rights, it may also create challenges for startups in terms of compliance costs and innovation.

The likelihood is we’ll see near immediate shifts in the venture world as any current potential investments will need to be assessed against the law regardless of whether it’s in place… yet.

The long-term impact will depend on how the AI Act is implemented and enforced, as well as how other jurisdictions respond to the EU's approach.

Note: The EU Law is 88k words. That's a lot of words. I/we/it spent much of a day pulling it apart, trying to understand it and looking for 'Legitimate Interest' style loopholes in it. There are five significant ones that I/we/it identified. Doing a dive into each. This is the fourth in the series and in many ways will be existential to many AI startups out there and could cause a significant chilling effect on the industry as a whole.

Research: This was all Claude. Sorry GPT, you weren't required.

Narrative: Jon 50% Claude 50% after much discussion.