Altman and AI regulation: who bears the risks when speed comes first
Sam Altman makes the case for the benefits of AI despite its risks. The political question is who decides what damage is acceptable and who is liable for it.
Sam Altman, head of OpenAI, has argued in a media interview for accepting certain risks in exchange for the benefits of artificial intelligence. The wording captures a real trade-off: technologies can improve research, productivity and access to knowledge, while at the same time making fraud, manipulation, data protection violations or security risks easier. The decisive political question is therefore not whether risks exist. It is who accepts them — and who bears the costs.
Benefits are not a free pass
With AI, benefits and harm are rarely distributed equally. Companies can earn profits through faster processes, while those affected carry faulty decisions, data misuse or job losses. A general promise to develop «responsibly» is therefore not enough. Clear requirements are needed for testing, documentation, incident reporting and independent oversight.
In a recent policy paper, OpenAI itself calls for binding, capability-based rules for particularly powerful AI. The Frontier Governance Framework names risks in areas such as cyberattacks, CBRN misuse, manipulation and loss of control. That position is more nuanced than a blanket rejection of regulation — but it also shows the tension between voluntary corporate frameworks and enforceable law.
What Switzerland can take from this
Switzerland is not a spectator. Companies and authorities use global AI services; at the same time, data protection law, sector-specific obligations and international standards apply. For small and medium-sized enterprises, rules that are easy to understand matter especially: they need legal certainty without only large corporations being able to handle the burden of testing and documentation.
A sensible framework separates risks by use and capability. A text tool for internal drafts demands something different from a system that decides on loans, medical priorities or critical infrastructure. Where consequences are hard to reverse, transparency, human accountability and independent channels for complaints must be stronger.
Progress does not become safer because risks are linguistically accepted. It becomes safer when the people affected by errors receive protection, information and genuine legal remedies.
Sources
OpenAI, AI Policy Window: https://openai.com/index/ai-policy-window/ OpenAI, Frontier Governance Framework: openai.com/index/openai-frontier-governance-framework… OpenAI, Principles: https://openai.com/index/our-principles/ Bild, interview reference: bild.de/politik/ausland-und-internationales/chef-von-openai…
