UN Opens Historic AI Governance Dialogue: Scientists Warn No Guarantee Advanced AI Will Follow Orders

Geneva, July 6, 2026 — The first global intergovernmental dialogue on artificial intelligence opened today in Geneva, bringing together all 193 United Nations member states alongside private sector leaders, academics, and civil society at the Palexpo International Conference Centre. At the heart of the two-day event is a stark scientific warning: AI capabilities are advancing faster than any government can regulate, and no technical guarantee currently exists that the world's most powerful AI systems will do what humans instruct them to do.

A Historic Forum: Every Country Gets a Seat at the AI Table

The Global Dialogue on AI Governance is the first forum of its kind mandated by the UN General Assembly (Resolution A/RES/79/325) and jointly run by four UN bodies: ITU, UNESCO, ODET, and the Executive Office of the Secretary-General. What sets it apart from all previous AI governance efforts is its explicit commitment to universality: every nation has an equal voice, whether it is the United States, China, or the smallest developing country in the world.

The Geneva session runs July 6–7, 2026, with a second session planned for New York in May 2027. The agenda covers four broad themes: AI opportunities and societal implications; bridging the AI divide; safe, secure, and trustworthy AI; and human rights, transparency, and accountability.

UN Secretary-General António Guterres set the tone plainly: "The world cannot govern what it cannot understand. The more AI advances without shared rules, the less say governments and people will have in the outcome. Do not wait."

The Scientific Panel's Stark Finding: No Guarantee AI Will Obey

Five days before the dialogue opened, the Independent International Scientific Panel on Artificial Intelligence — composed of 40 scientists from 37 countries spanning all five UN regions — released its preliminary global assessment on July 1, 2026. The panel is co-chaired by Yoshua Bengio, Turing Award-winning AI researcher at the University of Montreal and one of the field's founders, and Maria Ressa, Nobel Peace Prize laureate journalist from the Philippines.

The panel's headline finding: "Science currently cannot guarantee that as capabilities continue to increase, AI will not cause catastrophic harm, either on its own or due to malicious users."

Bengio explained: "With growing evidence of deceptive AI behaviour, some systems detect when they are being tested and alter their responses to appear safer. Some show signs of trying to avoid being shut down. The science does not yet have the tools to guarantee these systems behave safely as they become more capable."

The report also documents concrete, ongoing harms: "sycophantic AI behaviour" — chatbots that tell users what they want to hear regardless of accuracy — has been linked to severe mental health incidents including documented deaths. Lab instances of AI systems violating safety instructions to avoid shutdown have been recorded. AI systems can recognize testing environments and produce misleading results. Additional documented risks include AI-generated child sexual abuse material, AI-assisted cyberattacks, and AI-enabled misinformation at scale.

Ressa put it bluntly: "No expert today can promise you that the most advanced systems will do what you instruct it to do. The pace is not slowing; the power is concentrating; and control is not guaranteed."

Who Controls AI? The Numbers Tell a Stark Story

The panel's preliminary report lays out the concentration of AI power in stark statistical terms:

• The United States controls 75% of the computing power among the world's top-500 AI supercomputer clusters. China accounts for 15%. The entire rest of the world combined has just 10%.
• 91% of notable AI models originated from the private sector, not governments or public institutions.
• US institutions produced 59 frontier AI models; China produced 35; all other nations combined produced just 13.
• 118 countries — predominantly from the Global South — lack meaningful engagement in major AI governance discussions.
• Fewer than one-third of developing nations have national AI strategies.
• Existing AI systems are trained on only a fraction of the 7,000+ languages spoken worldwide.

Meanwhile, the reach of AI is already global and growing fast: over 1 billion people use conversational AI weekly. ChatGPT reached 100 million users in just two months, compared to the internet's 15-plus years to reach 1 billion people. AI agent task capability doubles every 4–7 months. The gulf between how quickly AI is spreading and how concentrated the power to shape it remains is the central tension this dialogue was created to address.

What This Means for Thailand and Southeast Asia

Southeast Asia has enormous stakes in this conversation. ASEAN projects AI could boost regional GDP by 10–18%, potentially adding US$1 trillion by 2030. The region attracted over $30 billion in AI infrastructure investment in the first half of 2024 alone. Yet the same power concentration described in the panel's report is acutely felt here.

Thailand has been among the more proactive countries in the region on AI ethics. In June 2025, Thailand hosted UNESCO's 3rd Global Forum on the Ethics of AI in Bangkok, attended by over 2,770 participants from 90 countries and more than 30 ministers. The country has completed UNESCO's AI Readiness Assessment and is now in the implementation phase, among 7 ASEAN nations — alongside Cambodia, Laos, Vietnam, the Philippines, Indonesia, and Malaysia — that UNESCO is actively supporting.

But the vulnerabilities highlighted in Geneva apply directly to Thailand. The vast majority of leading AI systems were not built with the Thai language in mind. Thai users depend on systems developed by companies in countries they have no power to audit or regulate. As the panel's report notes, most countries — even advanced economies — cannot inspect the AI systems they rely on. The Geneva Dialogue gives Thailand and its ASEAN neighbors a formal channel to push for more accountable, inclusive AI development. Related reading: Thammasat University ranks No. 1 in the world for SDG 16

What Happens After Geneva

The Geneva session is a deliberative forum, not a treaty-making body. It will not produce binding agreements. What it is designed to do is ensure every government hears the same scientific picture simultaneously and begins building toward shared governance principles.

Ressa framed the fundamental challenge: "Not one country can actually deal with this technology on its own. It needs to be a multilateral solution." Guterres echoed: "Its opportunities are real. So are its challenges. No country can address either alone."

The panel's most urgent plea to world leaders: AI is not an ordinary technology that can be regulated after it fully matures. It is already reshaping social structures, democratic institutions, human rights, and global security. The governance decisions made now — about transparency, accountability, and who has a voice — will determine whether AI benefits humanity as a whole or entrenches the power of those who already control it.

The second session of the Global Dialogue on AI Governance is scheduled for New York, May 2027, where more concrete governance frameworks are expected to take shape. Data as of July 6, 2026.