OpenAI and Anthropic leaders call for global cooperation for AI safety

Introduction: A defining moment for artificial intelligence governance

In a global context marked by the unprecedented acceleration of the development of artificial intelligence systems, two of the most influential leaders in the global tech industry have taken a significant step on the stage of international diplomacy. Sam Altman, CEO of OpenAI, and Dario Amodei, CEO of Anthropic, spoke before the United Nations, explicitly and eloquently calling for structured global cooperation on the safety of AI systems. This call is not a simple façade, but reflects a genuine and technically substantiated concern about the systemic risks that next-generation artificial intelligence models, especially AGI (Artificial General Intelligence) or frontier systems that already operate at superhuman levels of performance in specialized domains, may pose.

The context in which this intervention took place is extremely relevant: September 2026, a year in which the capabilities of AI models have advanced considerably from the standards known to the general public in 2024-2025. State-of-the-art multimodal models, autonomous reasoning systems, and AI agents capable of performing complex tasks with minimal supervision are now an operational reality, not a theoretical projection. In this landscape, the lack of an international regulatory framework represents a critical vulnerability at the global level.

What did Altman and Amodei ask for before the UN?

Main points of the joint speech

The two leaders argued that the development of AI at a global level can no longer be left exclusively in the hands of private corporations or individual governments, no matter how technologically advanced they may be. The central message was clear: AI is a technology with civilizational impact, and its risks must be managed through an international cooperation mechanism, similar to nuclear treaties or biological weapons agreements. This analogy is not chosen by chance — it signals the level of danger that the founders of the most advanced AI labs associate with the uncontrolled development of these systems.

through specific requests formulated before UN representatives include:

    Creating an international body dedicated to monitoring the development and deployment of frontier AI systems, with a real mandate for inspection and reporting.
    Establishing mandatory minimum safety standards for training models that exceed certain computational power thresholds (measured in FLOPs).
    Implementation of a global incident reporting system generated by AI systems, similar to existing mechanisms in international civil aviation.
    Cooperation between states to prevent the use of AI in the development of biological, chemical or cyber weapons with the potential for mass destruction.
    • Creating diplomatic channels dedicated to the exchange of technical information on the risk assessment of advanced AI models.

Technical positioning of arguments

From a technical point of view, the argument of the two executive directors was supported by concrete references to the evolution of AI model capabilities. Amodei, a former top researcher in the field of AI safety, pointed out that current large language models (LLMs) have already outperformed humans in numerous cognitive benchmarks, and agent-like AI systems—which can act autonomously, plan long-term, and interact with complex digital environments—raise alignment challenges that the industry has yet to fully address. The problem of alignment, or ensuring that an AI system acts in accordance with human values ​​and intentions, remains one of the most difficult challenges in modern computer science.

Altman completed This technical perspective is accompanied by an analysis of the speed of adoption: companies and governments are integrating AI systems into critical infrastructures — energy grids, healthcare systems, financial infrastructure, defense — at a pace that exceeds current risk assessment capacity. This asymmetry between the speed of deployment and the maturity of assessment mechanisms represents a systemic risk at a global level.

Why does international cooperation in the field of AI Safety matter?

The limits of national regulation

A fundamental aspect highlighted in the UN intervention is that National regulations, no matter how well-intentioned, are insufficient in the face of a cross-border technological phenomenon. An AI model trained in one jurisdiction can be accessed and used from anywhere in the world. A vulnerability discovered in an AI system deployed in Europe can be exploited by state or non-state actors in Asia or North America in the same timeframe. In the absence of a global coordination mechanism, states act in isolation, with fragmented and often incompatible regulations.

The European Union has taken important steps by AI Act, gradually taking effect starting in 2024, which classifies AI systems by risk level and imposes strict requirements for high-risk categories. The United States has issued executive orders and initiated regulatory processes through NIST and other federal agencies. China has implemented its own regulations on generative algorithms. But these initiatives remain isolated and uncoordinated at the global level, creating so-called regulatory arbitrage opportunities — situations where actors may choose more permissive jurisdictions to avoid stricter safety standards.

Specific technical risks justifying urgency

From a technical perspective, there are several categories of risks that justify the urgency of an international framework:

Biosecurity issues: Advanced AI models can significantly speed up the design of pathogens or toxins, reducing the technical barriers traditionally required for malicious actors. This risk has been flagged by researchers at MIT, Johns Hopkins, and other leading academic institutions.

Autonomous cyber attacks: AI systems can generate and execute cyberattack campaigns at a speed and scale that exceed human capacity for real-time response, endangering national and international critical infrastructures.

Disinformation on an industrial scale: Generative models can produce fake, personalized, and highly persuasive content in massive volumes, with the potential to destabilize democratic processes and social cohesion.

Concentration of power: Asymmetric access to advanced AI technologies can create profound geopolitical imbalances, in which states or corporations that control the most advanced models gain disproportionate strategic advantages in economics, defense, and diplomacy.

Long-term existential risks: Although controversial in academia, scenarios involving AI systems that exceed human control are a concern taken seriously by researchers such as those at the AI ​​Safety centers at Oxford, Cambridge, and UC Berkeley.

The international community's reaction and implementation challenges

How did the diplomatic community respond?

Altman and Amodei's joint appeal was received with a mixture of sincere interest and pragmatic skepticism by representatives of UN member states. G7 countries have generally expressed support for initiating formal discussions, while some emerging states have raised legitimate concerns about the fairness of a potential global framework: who will set the standards? Who will control the monitoring body? Will smaller economies be disadvantaged in accessing AI technologies if the standards are too restrictive?

These questions are not rhetorical — they reflect real tensions in the geopolitical dynamics of technology. There is a real risk that a potential international AI regulatory body would be dominated by the interests of major technological powers, replicating the inequalities that exist in other international forums. Ensuring equitable representation and democratic access to the benefits of AI it must be an integral part of any global cooperation framework to be truly legitimate and effective.

Institutional precedents and possible models

The two CEOs' UN initiative is part of a broader trend of searching for institutional models for global governance of AI. Among them models discussed in policy circles and think tanks are:

IAEA (International Atomic Energy Agency) model: A body with a mandate to inspect and verify the compliance of member states with safety standards, adapted to the specifics of AI.

ICAO (International Civil Aviation Organization) model: Global technical standardization and incident data exchange, applied to AI systems in critical infrastructures.

The IPCC (Intergovernmental Panel on Climate Change) model: A forum for scientific synthesis and risk assessment, which informs political decisions without having direct regulatory power.

Each of these models presents specific advantages and limitations when transposed into the AI ​​context. The speed of evolution of AI technology is significantly higher than any previously internationally regulated field, which means that any governing body must have the ability to adapt quickly to major technical changes, without being paralyzed by bureaucracy or geopolitical disagreements.

Implications for the AI ​​industry and research labs

Corporate responsibility in the global AI ecosystem

The fact that the leaders of two of the most important companies in the frontier AI field are taking the initiative to call for stricter regulation is, in itself, a strong signal. This positioning is not without strategic interest — large companies can benefit from regulations that raise barriers to entry for smaller competitors — but it also reflects a genuine awareness of the responsibility that the labs that develop the world’s most powerful AI systems have. Both OpenAI and Anthropic have invested heavily in dedicated AI Safety and Alignment teams, publishing research in areas such as constitutional AI, reinforcement learning from human feedback (RLHF), interpretability and red-teaming.

At the same time, the UN initiative is putting pressure on other major industry players — Google DeepMind, Meta AI, xAI, Mistral, and others — to adopt clear public positions on their commitment to safety standards and to participate constructively in global governance processes. Absence from this conversation could, in itself, become a significant reputational and regulatory risk. in the context where states are starting to legislate more aggressively in the field.

What does this mean for AI professionals?

For engineers, researchers, product managers, and all professionals working with or around AI systems, developments in the governance space have direct practical implications. Safety standards, auditability requirements, transparency and documentation obligations of models will increasingly become part of daily workflows, not just abstract concerns of legal or compliance departments. Understanding the fundamental concepts of AI Safety, risk assessment techniques and applicable regulatory frameworks is becoming an essential competency for any professional aspiring to a relevant role in the AI ​​industry of the coming years.

In addition, the demand for specialists in intersecting fields — AI Policy, AI Ethics, AI Auditing, Responsible AI Development — is growing rapidly, both in the private sector and in public institutions and international organizations. These are roles that combine solid technical knowledge with an understanding of the legal, ethical and geopolitical context, and which will be decisive in shaping how AI is developed and deployed in the next decade.

Conclusion: A call to action at the right time

Sam Altman and Dario Amodei's speech at the UN in September 2026 marks a moment of institutional maturity for the AI ​​industry. Public recognition, even by the architects of the world's most advanced AI systems, of the need for a global safety framework is an important catalyst for conversations that should have happened years ago. The challenges are immense — from the technical complexity of evaluating AI systems, to geopolitical tensions between major powers, to the risk of regulation stifling beneficial innovation — but the absence of coordinated action is, in itself, a decision with serious consequences.

The next period will be critical: whether the UN appeal will generate concrete institutional mechanisms or will it remain a symbolic gesture depends on the political will of Member States, the commitment of industry and the pressure of civil society. What is certain is that artificial intelligence is no longer a technology of the distant future — it is a present reality, with a measurable impact on the global economy, security and social cohesion, and its responsible management is one of the most important challenges of our generation.

You have certainly understood what is new in 2026 related to artificial intelligence. If you are interested in deepening your knowledge in the field, we invite you to explore our range of courses structured by roles and categories in AI HUBWhether you're just starting out or want to brush up on your skills, we have a course for you.