Southeast Asian countries are taking a business-friendly approach to artificial intelligence regulation in a setback to the European Union's push for globally harmonised rules that align with its own stringent framework.

Reuters reviewed a confidential draft of the 10-member Association of Southeast Asian Nations' (ASEAN) "guide to AI ethics and governance," whose content has not previously been reported.

Three sources told Reuters the draft is being circulated to technology companies for feedback and is expected to be finalised at the end of January 2024 during the ASEAN Digital Ministers Meeting. Companies that have received it include Meta META.O, IBM IBM.N, and Google GOOGL.O.

EU officials earlier this year toured Asian countries in a bid to convince governments in the region to follow its lead in adopting new AI rules for tech firms that include disclosure of copyrighted and AI-generated content.

In contrast to the EU's AI Act, the ASEAN "AI guide" asks companies to take countries' cultural differences into consideration and doesn’t prescribe unacceptable risk categories, according to the current version reviewed. Like all ASEAN policies, it is voluntary and is meant to guide domestic regulations.

With almost 700 million people and over a thousand ethnic groups and cultures, Southeast Asian countries have widely divergent rules governing censorship, misinformation, public content and hate speech that would likely affect AI regulation. Thailand, for example, has laws against criticising its monarchy.

Technology executives say ASEAN's relatively hands-off approach is more business-friendly as it limits the compliance burden in a region where existing local laws are already complex and allows for more innovation.

"We are also pleased to see this guide aligns closely with other leading AI frameworks, such as the United States’ NIST AI Risk Management Framework," IBM Asia's vice president of government affairs Stephen Braim said, referring to voluntary guidelines developed by the U.S. Department of Commerce's National Institute of Standards and Technology.

Meta and Google did not respond to requests for comment.

BENEFITS VS HARM

The guide, which is meant to be periodically reviewed, urges governments to aid companies through research and development funding and sets up an ASEAN digital ministers working group on AI implementation.

Senior officials in three ASEAN countries said they are bullish on the potential of AI for Southeast Asia and believe the EU has been too quick to push for regulation before the harms and benefits of the technology are fully understood.

The ASEAN guide advises companies to put in place an AI risk assessment structure and AI governance training, but leaves specifics to companies and local regulators.

"We see it as putting ‘guardrails’ for safer AI," one official told Reuters. "We still want innovation."

The guide warns of the risks of AI being used for misinformation, "deepfakes", and impersonation, but leaves it to individual countries to work out the best way to respond.

Other Asian nations such as Japan and South Korea have flagged similarly relaxed approaches to AI regulation, casting doubts over the EU’s ambition to establish a global standard for AI governance based on the rules that would apply to its 27 member states.

Driving the EU push are concerns in Brussels about the rapid pace of AI development and its effect on civil rights and security, which have put risk controls and enforcement at the centre of the proposed legislation.

While ASEAN does not have any powers to make laws, its preference for member states to make their own policy determinations puts those countries on a distinctly different track to the EU.

The EU's struggles to create global consensus on AI regulation contrast with its mostly successful campaign last decade to establish data protection laws that have become a template for other major economies around the world.

"What we think is important is to have similar principles," a European Commission spokesperson told Reuters. "We are not seeking full harmonisation, as we are mindful of cultural differences, however, we regard the underlying principles as important."

EU officials and lawmakers told Reuters that the bloc would continue to hold talks with Southeast Asian states to align over broader principles.

"If we want AI to be used for good, we need to come together on the basic principles of human rights," Dutch minister for digitalisation Alexandra van Huffelen told Reuters. "I don’t think we are very far off from that we couldn’t bridge the differences."