Who governs the advancing AI?
© Imagesinesââistock/Getty Images
Whether youâre excited about artificial intelligence (AI), frightened by it, or a little bit of bothâyou may be wondering how AI is regulated, and whether regulators are keeping up with the rapid pace of advancement. Regulation matters for all stakeholders of an AI system, whether youâre an investor, entrepreneur, AI developer, or consumer engaging with an AI platform.
Although there are regulations out there that cover data privacy, data protection, and intellectual property rights, when a technology like generative AI bursts onto the scene and begins rapid advancement and adoption, itâs hard for regulators to keep pace.
Plus, your geographical location matters for AI regulation, even though many AI tools are in use worldwide. Hereâs a rundown of AI regulations, as well as the pros and cons of governing this powerful technology.
Key Points
- Laws concerning the use of personal data are the most advanced, but AI models need more transparency.
- Rules about bias prevention and auditing can produce more equitable outcomes.
- The counterarguments to enhanced regulation include concerns over stifling innovation, the costs of compliance, and the potential speed of obsolescence.
Laws and regulations specifically governing AI are pretty scarce in 2024âunless the AI model uses personal data that isnât anonymized. AI systems in health care, finance, insurance, and lending are among the sectors that use personal data to make customized recommendations.
Letâs examine the most important laws and regulations that govern AI models using personal information.
1. General Data Protection Regulation
In effect in the European Union, the General Data Protection Regulation (GDPR) is an influential regulatory framework billed as the âstrongest privacy and security law in the world.â The GDPR broadly governs collecting, storing, and processing personal data.
Some key highlights of the GDPR that impact AI models include:
- Individuals must provide âclear consentâ to process personal data, must be given the right to object to having their data used, and must have the right to erasure of their personal data.
- Enterprises are required to implement appropriate data security measures.
- Transfer of personal data to non-EU countries is covered by the GDPR.
2. California Consumer Privacy Act
Passed in 2018 and effective as of 2020, the California Consumer Privacy Act (CCPA) aims to give individuals more control over the personal information that businesses collect about them. The Act defines individualsâ rights and establishes certain requirements for businesses that conduct business in California.
All of these rights established by the CCPA may be relevant to AI models:
- The right to know about the personal information that a business collects.
- The right to delete personal information collected.
- The right to opt out of the sale or sharing of personal information.
- The right to correct inaccurate personal information.
3. Personal Information Protection and Electronic Documents Act
Canada has had a data privacy law called the Personal Information Protection and Electronic Documents Act (PIPEDA) on the books since 2000. The law applies to âprivate-sector organizations across Canada that collect, use, or disclose personal information in the course of a commercial activity,â according to the Office of the Privacy Commissioner of Canada.
PIPEDA matters for AI systems, as it:
- Requires organizations to obtain individualsâ consent to collect, use, or disclose personal information.
- Restricts the use of personal information to only the purposes for which it was collected.
- Applies to all businesses that operate in Canada and handle personal information that crosses provincial or national borders.
4. General Data Protection Law
Brazil in 2018 passed the General Data Protection Law to establish data processing rules and personal data protections to safeguard individualsâ privacy.
Whatâs potentially relevant about the General Data Protection Law for AI systems:
- Data processing is permissible only for legitimate, specific, and explicit purposes of which the data subject (the individual) is informed.
- Information about data processing activities must be clear, precise, and easily accessible.
- The law applies to data processing operations in Brazil regardless of where the data processor is located.
What makes AI regulation challenging
So, thereâs a regulatory structure in place across the globe to ensure data privacy and protection. Problem solved, right? Not so fast. AI has changed the game, so to speak.
- AI technology is evolving rapidlyâtypically faster than regulatory frameworks can adapt.
- The complexity and diversity of AI systems make it difficult to establish uniform regulations.
- The global nature of AI development and deployment creates immense jurisdictional complexity.
- AI lacks universally accepted standards for evaluating and certifying the technology.
Despite these challenges, itâs critical to advance the regulation of artificial intelligence. AI used improperly, especially by enterprises and governments, could produce many unwanted effects.
What happens if AI is overregulated?
Pros and cons of regulating AI
Well-regulated artificial intelligence is likely to provide more benefits to more people, but that doesnât mean that AI regulation has zero drawbacks. The pros and cons show how complicated the issue is.
Pros of regulating AI:
- Increases AI model transparency. Regulations may boost AI system transparency by requiring detailed disclosures about how the technology operates. For example, itâs unclear to what extent generative AI models use copyrighted material or protected data in their training.
- Enhances data privacy and protection. Widely regulated AI, regardless of the type of data used, would likely increase usersâ control over their data and the dataâs security.
- Fights bias in algorithms. Another potential positive effect of regulation is that it may reduce or eliminate bias in AI algorithms. Rules about bias prevention and auditing can produce more equitable outcomes.
Cons of regulating AI:
- May stifle innovation. Overly stringent or premature AI regulations can hinder the development and adoption of AI by slowing research and experimentation. In an extreme case, overly burdensome regulation in developed nations could lead to so-called regulatory arbitrage, whereby AI developers choose to operate in areas of the world with a more favorable regulatory structure.
- Can create an economic burden. Understanding and complying with AI regulations can impose significant costs on businesses, especially early-stage start-ups. Regulation may create an economic barrier to entry.
- Rules can become outdated. Lawmakers are challenged to craft regulations that are specific enough to address the unique challenges of AI, yet flexible enough to remain relevant as the technology evolves.
The bottom line
With great power comes great responsibility. Thatâs true for AI systems developers, and true for the governments that regulate them. Regulators wishing to apply more developed and cohesive standards to the AI industry are responsible for devising rules that simultaneously protect users and foster innovation.
Thatâs a delicate balance to strikeâand the outcome is important. Intentionally or otherwise, regulationâor the lack thereofâwill pull all stakeholders into the ethics of artificial intelligence.
This article was most recently revised and updated by Allie Grace GarnettThis article was most recently revised and updated by Doug Ashburn
