With the rapid development of powerful artificial intelligence services like ChatGPT, regulators are resorting to existing regulations to control a technology that has the potential to reshape societies and businesses.The European Union is leading the way in drafting new AI rules that could serve as a global benchmark, addressing privacy and safety concerns that have arisen with the rapid advancement of generative AI technology, such as OpenAI's ChatGPT.
However, the enforcement of these regulations is expected to take several years.
"In the absence of specific regulations, governments can only apply existing rules," said Massimiliano Cimnaghi, a European data governance expert at consultancy BIP. "Data protection laws are applied to protect personal data, and regulations that have not been specifically defined for AI but are still applicable come into play when there is a threat to people's safety."
In April, Europe's national privacy watchdogs established a task force to address concerns related to ChatGPT following the Italian regulator Garante's temporary shutdown of the service. Garante accused OpenAI of violating the EU's General Data Protection Regulation (GDPR), a comprehensive privacy regime implemented in 2018.
ChatGPT was reinstated after OpenAI agreed to incorporate age verification features and allow European users to block the use of their information for training the AI model.
A source close to Garante revealed that the agency plans to extend its examination to other generative AI tools. Additionally, data protection authorities in France and Spain have initiated probes into OpenAI's compliance with privacy laws.
Bringing in AI experts is a priority for regulators. Generative AI models have gained notoriety for producing errors or "hallucinations," generating misinformation with surprising confidence. Such errors could have significant consequences, leading to unfair rejections for loans or benefit payments if AI is used by banks or government departments to expedite decision-making processes. Major tech companies like Alphabet's Google (GOOGL.O) and Microsoft Corp (MSFT.O) have already ceased using AI products with ethical concerns, particularly in the financial sector.
Regulators aim to apply existing rules that cover various aspects, including copyright, data privacy, and two key issues: the data used to train AI models and the content they generate. Experts and regulators from the United States and Europe highlight the importance of agencies "interpreting and reinterpreting their mandates." For example, the US Federal Trade Commission (FTC) is investigating algorithms for discriminatory practices under its existing regulatory powers.
In the European Union, proposals for the AI Act will require companies like OpenAI to disclose any copyrighted material used, such as books or photographs, to train their models, potentially exposing them to legal challenges. However, proving copyright infringement may not be straightforward, as lawmakers acknowledge the complexity involved.
French data regulator CNIL is taking a "creative" approach to examine how existing laws can be applied to AI. In France, discrimination claims are typically handled by the Defenseur des Droits (Defender of Rights). However, CNIL has taken the lead on AI bias due to the Defender of Rights' limited expertise in this area. While data protection and privacy remain their main focus, CNIL is exploring the full range of effects and considering using GDPR provisions that protect individuals from automated decision-making. However, reaching a consensus among regulators may prove challenging, with potential differences in views and approaches.
In the UK, the Financial Conduct Authority, among other state regulators, is developing new guidelines for AI. They are collaborating with the Alan Turing Institute in London, as well as legal and academic institutions, to enhance their understanding of the technology.
No comments:
Post a Comment