We need guidelines that are not merely regulatory but architectural.
The accelerating evolution of artificial intelligence demands that we revisit the foundations of our shared human compact. In my 27 July article in these pages, I argued that the international community must anchor AI governance in universal norms rather than in fragmented national approaches. Today, as artificial generative intelligence (AGI) emerges as a distinct and potent subset of frontier AI, that imperative becomes even more urgent.
Artificial generative intelligence—defined here as systems capable of autonomously producing novel text, images, code, and multimodal content—sits within the broader category of frontier AI, the most advanced and compute-intensive models now achievable. Frontier AI encompasses the full spectrum of high-capability systems, including reasoning engines, autonomous agents, and multimodal interpreters. AGI is the generative branch of this frontier: the part that creates, imitates, and increasingly initiates content.
It is precisely this generative capacity that places AGI at the crossroads of innovation and human rights.
The Universal Declaration of Human Rights (UDHR), adopted in 1948, remains the world’s most widely accepted articulation of human dignity. Its articles are not relics; they are living commitments. Several of them speak directly—indeed presciently—to the challenges posed by AGI.
Article 1, affirming that all human beings are “endowed with reason and conscience,” reminds us that AGI must never be mistaken for a bearer of human moral agency. As generative systems grow more persuasive and more autonomous, engineering constraints must ensure that they do not simulate moral authority or impersonate human conscience.
Article 3, guaranteeing the right to life, liberty, and security of person, demands that AGI development incorporate safety-by-design. Generative systems capable of producing harmful biological, chemical, or cyber-technical content must be bounded by strict access controls, monitored compute thresholds, and real-time misuse detection.
Article 12, protecting individuals from arbitrary interference with privacy, is directly implicated by AGI’s appetite for data. Generative models trained on vast human datasets must be engineered to avoid re-identification, prevent leakage of personal information, and maintain strict provenance controls. Privacy cannot be an afterthought; it must be a design constraint.
Article 18, safeguarding freedom of thought, becomes newly relevant as AGI systems generate persuasive political, ideological, or religious content. Guardrails must prevent generative engines from shaping belief through covert influence, targeted manipulation, or synthetic consensus.
Article 19, guaranteeing freedom of expression, requires a delicate balance. AGI can amplify human creativity, but it can also drown it. Engineering constraints should ensure that generative systems label synthetic content clearly, preserve the authenticity of human speech, and avoid overwhelming public discourse with machine-generated noise.
Article 27, affirming the right to participate in cultural life, speaks to AGI’s impact on creativity itself. Generative systems must respect intellectual property, protect artistic integrity, and avoid appropriating cultural expressions without consent.
Taken together, these articles form a blueprint for responsible AGI evolution. They suggest guidelines that are not merely regulatory but architectural:
• Transparency of origin for all generative outputs.
• Provenance tracking for training data.
• Capability thresholds that trigger mandatory safety evaluations.
• Human-in-the-loop requirements for high-risk generative actions.
• Compute governance to prevent uncontrolled scaling.
• Identity safeguards to prevent impersonation of individuals or institutions.
The UDHR does not oppose innovation; it guides it. Artificial generative intelligence will shape the next era of human creativity, communication, and cognition. But it must do so in a way that respects the dignity of the human person, preserves the integrity of human agency, and strengthens—not supplants—the universal norms that bind us together.
As AGI advances, our task is clear: to ensure that the machines we build never outrun the values that define us.
*Hugh Dugan, President of Multilateral Accountability Associates and former White House Senior Director in the National Security Council.
Part II to author’s 27 July article on AI and global norms
