TRAIF 2026 Impulse Session: What Are We Missing in the Responsible AI Conversations?

Panelists discussed missing voices, power imbalances and governance gaps in the global responsible AI conversation.

The Responsible AI Forum (TRAIF) 2026 opened with an ambitious question. Held on May 19th, 2026 at Amerikahaus in Munich, the Impulse Session, titled “What Are We Missing in Responsible AI Conversations?”, brought together scholars and practitioners to examine the blind spots of a field that often speaks confidently about values it has yet to fully examine. Moderated by Dr. Auxane Boch (IEAI, TUM), panel participants included Prof. Christoph Lütge (IEAI, TUM), Prof. Mohammed Khalil (Hassan II University; Morocco AI), Prof. Anne-Elisabeth Courrier (Nantes University; Emory Center for Ethics), and Prof. Thomas Maak (University of Queensland).

Whose Values Are We Talking About?

Prof. Khalil began by stepping back from familiar keywords such as “fairness”, “transparency” and “ethics”, and instead took a historical approach. While humanity has made extraordinary progress in science, technology and engineering, he argued, the same cannot be said for the humanities, social norms, inclusion and diversity. To illustrate this, he pointed to the University of al-Qarawiyyin in Fez, Morocco, the oldest continuously operating university in the world, founded over a thousand years ago. What makes it remarkable, he noted, is not its age but the fact that its founder was a woman, Fatima al-Fihri, and that society at the time accepted her as a leader. He asked the audience to consider how many women today serve as CEOs of AI companies or lead major research institutions, drawing a direct line between that historical precedent and the present underrepresentation of women in AI leadership.

Prof. Khalil then turned to the global dimension of the responsible AI conversation. Global North frameworks, he observed, naturally emphasize individual values such as privacy and personal rights, while Global South perspectives center on community, family, collective responsibility and solidarity, values captured in the African philosophy of Ubuntu: “I am because we are”. Rather than defending one side over the other, he urged a shift in framing: “Let’s not talk about Global North or Global South. Let’s talk about the global world”. He stressed that the world must be understood as a single body; if any part of the system suffers under AI deployment, the system as a whole has failed.

On how to fix it, Prof. Khalil called for alignment with plurality: “We cannot build 21st-century AI on 20th-century social norms, and the field must develop a genuinely global perspective in which every culture finds its place”. He invoked the Elephant Clock of Al-Jazari as a symbol of this vision, a machine that brought together Indian, Chinese, Persian and Greek civilizations in a single object, and argued that the question must shift from who will win the AI race to what will be the greatest benefit for humanity. Sovereignty matters, he stressed, but it must not mean isolation: “We are all on the same boat, and we have to succeed together”.

Asymmetry in Space and Time

Prof. Courrier noted that the discussion on responsible AI is based on a false assumption that we live in a flat world with common global standards for what responsible AI is, what harm is and who is responsible. The blind spot, she suggested, is asymmetry in two dimensions. The first is spatial: we do not see data with the same eyes, we do not share the same standards or values and our regulatory capacities vary enormously across regions. The second is temporal: AI is accelerating decision-making, production, education and the relation to power at a pace that democratic institutions simply cannot match. “For AI, five years is enormous. For a parliament, it’s nothing.”, she noted. She argued that the EU AI Act was already struggling to remain relevant upon publication, as agentic AI systems emerged that its drafters had not anticipated. Prof. Courrier called for adaptive governance-frameworks that can be continuously revised rather than fixed in place. She also stressed the need to include students and younger generations in governance processes, as those most affected by AI’s longterm consequences are underrepresented in decision-making.

On how to address these asymmetries, Prof. Courrier stressed that the challenge is political rather than technical and called for productive disagreement and adaptive governance. She also made the case for changing the questions we’re asking as responsible AI isn’t just about the ethics of the technology but about the larger ecosystem that surrounds it, who has the infrastructure, who owns the data and who has the capacity to do audits. “So long as operational questions are kept out of the mainstream discussion, the field will miss the point.”

Power, Accountability and the Missing Checks

Prof. Maak began by stating that AI is biased and will remain so. It is not neutral and cannot be neutral, because it reflects the people who create it. The AI workforce is about 85 percent male. And decisions are made by a small group of developers and executives. He looked closer at who is making those decisions, an “interesting group”, young billionaires, many in their twenties, thirties and forties, making decisions for the whole of humanity, mainly motivated by the desire to be first and to define the technology of the future. Some in the industry have quasi-religious views on technological submission, and a fringe element promotes ideas that border on techno-authoritarianism. The hyper-competitive dynamic has produced murky motivations, and the checks and balances to question them are mostly absent.

The extent of the power shift exacerbates the problem. NVIDIA’s market capitalization now stands at $5 trillion, surpassing the entire German economy, and five of the fifteen largest economies in the world are technology companies rather than nation-states, a share that has grown from roughly half to 70% over the past 15 years. The problem, Prof. Maak said, is not a lack of principles but a lack of mechanisms to hold decision-makers accountable. A responsible agency isn’t seen at the corporate level, precisely because hyper-competition means nobody wants to be left behind. He cited Dario Amodei at Anthropic as a rare exception, characterizing his approach as “intelligent self-binding”, the voluntary restriction on some military applications of the company’s technology. He identified deliberative democracy as the most promising avenue, grounding this in Jürgen Habermas’s work on discourse ethics, and called for cross-sector dialog among diverse stakeholder groups willing to engage with critical questions about the purpose and future of technology. Drawing on indigenous thinking, Prof. Maak highlighted its emphasis on long-term stewardship across generations and on relationality and community as correctives to the short-termism of the tech sector. He concluded by observing that OpenAI was founded with a humanistic mission that has since been overtaken by commercial pressures and revitalizing that dimension is one of the most urgent tasks ahead. When Dr. Boch raised the question of humility, Prof. Maak agreed, adding it is the industry’s shared ethos that it is acceptable to break things, which is dangerous when applied to highly capable technology, even if creative destruction has its legitimate place.

Perspectives We Are Missing

Prof. Lütge pushed back gently on Prof. Maak’s framing, stating from a business ethics perspective, something can be both for good and for profit. What is missing from many conversations, he argued, is a recognition that AI fundamentally challenges perspectives, not only in underrepresented regions but also in countries that consider themselves well-positioned and in no need of change. He named Germany specifically. At the political level, there is still insufficient recognition of the fundamental shift underway, as illustrated by Germany’s excellence initiative-of 70 research clusters funded at approximately 30 million euros each, only two focus on AI, and even with a generous count, the total reaches no more than four.

Prof. Lütge pointed to the Institute for Ethics in AI’s work with automotive engineers on autonomous driving as a model of how responsible AI needs to be integrated within the development teams as well as with engineers and product managers, and not just stay in the high-level policy discussions. When addressing power imbalances, he argued for a genuinely intercultural perspective. At the IEAI, this has meant engaging not just with Kantian ethics or utilitarianism but also with Ubuntu ethics and sustained dialog with research centers across Asia, the Americas and elsewhere. The real work involves understanding what different ethical perspectives actually mean in practice when applied to concrete AI systems. This was evident during the longer session discussion, where Prof. Khalil had made the point that sovereignty cannot mean isolation, and the race to lead in AI cannot be at the expense of global collaboration.

Q&A: Psychological Impact, Agentic AI and Ethics Washing

An audience member raised the psychological impact of AI on individuals, noting that despite the field’s growing body of principles, the human being is consistently overlooked in responsible AI discourse. Prof. Maak agreed, pointing to grief bots and romantic AI as examples of applications with poorly understood psychological consequences. Dr. Boch noted that the field produces a great deal of commentary on the issue, but very little concrete data, and that no adequate framework yet exists to systematically measure these effects. Prof. Courrier added that the topic is genuinely hard to sell to institutions noting that at the Atlanta AI Week, she spent two months advocating for a panel on the human side of performance in an AI-driven world, covering information overload, accelerating decision-making, and what a Microsoft report described as the “infinite workday”.

A second question addressed the governance of agentic AI systems. Prof. Courrier linked this to temporal asymmetry, noting that with agentic AI, development now moves on a monthly rather than yearly basis, making adaptive governance a necessity. Prof. Khalil cautioned against over-reliance on shifting terminology, as AI has passed through predictive, generative and now agentic phases, but the underlying philosophical questions remain constant. “There were only ten pages on neural networks in Stuart Russell’s foundational AI textbook, but almost all subsequent development is based on it”, he said. “The bigger debate may be whether the industry moves towards world models rather than continuing to train systems on synthetic data, which experiments suggest degrades in quality over time”, Prof. Maak said.

A further question addressed the current state of ethics washing in both corporate and academic AI discourse. Prof. Maak observed there is currently little traction in the field and argued that what is needed is not a new theory of AI but a reignition of the debate at the interface of existential risk and ethics-who should be involved, where limits should be set and when development should pause. Prof. Lütge noted much of the relevant work is already happening in computer science and technical disciplines, and pointed to the 2024 World Congress of Philosophy in Rome, where an AI ethics section appeared for the first time as evidence of growing engagement.

The final question addressed accountability when AI systems act autonomously. Prof. Maak pointed to emerging frameworks around shared and distributed agency as a basis for allocating responsibility in hybrid human-AI decision-making. Prof. Khalil noted accountability already operates across all stages of an AI product’s lifecycle, with global efforts underway to assign responsibility at each step. Prof. Courrier observed that while shared moral responsibility may be achievable, legal accountability remains considerably more difficult to establish. Prof. Lütge emphasized organizational responsibility and clear redress mechanisms as the most pressing practical priorities.

Closing Remarks

The session closed with a discussion on responsibility when AI acts autonomously. Panelists agreed responsibility will be distributed across stakeholders at every stage of a system’s lifecycle and legal and organizational frameworks for enforcing this accountability are still being developed. The IEAI thanked Amerikahaus for co-hosting and looks forward to continuing the dialog at future TRAIF events. The TRAIF 2026 Impulse Session took place on May 19th, 2026 in Munich.

The Responsible AI Forum 2026 was funded by the European Union. More on alignAI.

Watch a video of the panel below.