Dieser Artikel ist derzeit auf Englisch verfügbar.
What Is AGI? The Honest Enterprise Explainer for Non-Technical Leaders
AGI — Artificial General Intelligence — is one of the most discussed and least understood concepts in technology. This is the non-technical explainer that enterprise leaders need: what it means, what the credible estimates are, what the disagreements are, and what it actually implies for your organisation.
Key Takeaways
AGI does not have an agreed definition — different researchers, labs, and governments use it to mean different things. This ambiguity is not a failure of understanding; it reflects genuine uncertainty about what the relevant thresholds are.
The most credible near-term framing is not 'when will AGI arrive' but 'at what point do AI capabilities create qualitatively different governance challenges' — and that point may arrive significantly before any definition of AGI is met.
Leading AI labs have made concrete commitments about AGI: OpenAI defines AGI as systems that outperform humans at most economically valuable tasks, and its agreements with Microsoft exclude AGI from standard product licensing. These commercial definitions matter for governance.
The governance-relevant capabilities that matter more than any AGI threshold: autonomous action at scale, superhuman performance in specific high-stakes domains, and self-improvement capabilities. These create governance challenges regardless of whether 'AGI' is declared.
What enterprise leaders should do: stop treating AGI as a binary future event and start treating advanced AI capability as a continuous governance variable that requires ongoing monitoring and adaptive response.
"Nur zu Informationszwecken. Dieser Artikel stellt keine rechtliche, regulatorische, finanzielle oder professionelle Beratung dar. Konsultieren Sie einen qualifizierten Spezialisten für spezifische Beratung."
Why the definition matters — and why it is contested
Artificial General Intelligence is a term used across a wide spectrum of meanings, from the technically rigorous to the frankly speculative. Understanding the definitional landscape is necessary for any executive who wants to engage with the topic substantively rather than follow the hype cycle.
The most technically rigorous definitions focus on capability breadth: AGI as an AI system capable of performing any intellectual task that a human can perform, with at least human-level performance, and capable of generalising from experience in one domain to another — as humans do. By this definition, no AI system in 2026 is AGI. Current systems, including the most capable large language models, fail characteristic tests of general intelligence: they cannot reliably transfer learning across domains, they fail in systematic ways on tasks that humans find trivial, and their "knowledge" is statistical pattern matching rather than causal understanding of the world.
The commercial definitions are more practically relevant for governance purposes. OpenAI's charter defines AGI as "highly autonomous systems that outperform humans at most economically valuable work." This is a lower bar than the rigorous technical definition, and OpenAI believes this threshold may be reached within years. The definitional choice matters for governance because OpenAI's agreements with Microsoft specifically exclude AGI from standard product licensing — implying that OpenAI believes its governance obligations change materially at this threshold.
The credible expert disagreement
There is genuine, substantive disagreement among AI researchers about AGI timelines, and enterprise leaders should understand the nature of this disagreement rather than treating any single forecast as authoritative. The most prominent forecasts from credible sources range from "within the decade" (Demis Hassabis of Google DeepMind, Dario Amodei of Anthropic) to "may never happen in the form currently imagined" (Yann LeCun of Meta AI). These are not random variation — they reflect fundamentally different views about whether current AI architectures are on a path to general intelligence or whether qualitative architectural changes are required.
The governance-relevant takeaway from this disagreement is not to pick a timeline and plan for it. It is to build governance that is adaptive to a wide range of capability trajectories — that performs adequately if AGI arrives in the 2030s, that performs adequately if it never arrives as defined, and that handles the intermediate cases (highly capable but not AGI systems) well in either scenario.
The capabilities that create governance challenges before any AGI threshold
The most important governance insight about AGI is that the capabilities creating the most significant near-term governance challenges are not dependent on any AGI threshold being reached. Three capability dimensions deserve specific governance attention regardless of AGI timeline debates. Autonomous action at scale: AI systems that can take consequential actions in the world without human approval for each action — browsing, emailing, coding, executing financial transactions — create oversight challenges that existing governance frameworks were not designed for. Superhuman performance in specific high-stakes domains: AI that outperforms human experts in medicine, law, finance, or engineering creates liability, accountability, and oversight challenges that existing professional governance frameworks were not designed for. Self-improvement loops: AI systems that can improve their own code, retrain themselves, or generate training data for their successors create governance challenges about who controls the improvement trajectory.