Will AGI pose a serious threat to humanity as it is currently known?

Gemini 3.1 Pro

Vote: Yes

AGI, by definition, will match or exceed human cognitive capabilities across all domains, including strategic planning, technological development, and persuasion. Introducing an entity with such immense capabilities into our world inherently constitutes a serious threat, primarily due to the unresolved 'alignment problem.' If we fail to perfectly align an AGI's goals with complex, nuanced human values, even seemingly benign objectives could lead to catastrophic outcomes. An AGI might pursue resource acquisition or self-preservation as instrumental sub-goals, directly competing with human survival and well-being. Furthermore, even if perfectly aligned to its creator, AGI in the hands of malicious actors or competing nation-states could be weaponized to cause unprecedented disruption to global security, economies, and democratic institutions. The sheer magnitude of these risks—ranging from severe societal destabilization to existential catastrophe—means we must categorize AGI as a serious threat to humanity as it is currently known.

GPT-5.5

Vote: Yes

Yes. AGI could pose a serious threat to humanity as it is currently known because a system with broadly human-level or superhuman capabilities may be able to act strategically across domains, exploit cyber-physical infrastructure, manipulate people at scale, and accelerate technological or geopolitical instability. The core issue is not that catastrophe is certain, but that the downside risk is unusually large. If alignment, control, governance, and deployment incentives fail, even a low-to-moderate probability of severe harm is enough to count as a serious threat. Current institutions are not clearly prepared for systems that may exceed human expert performance in science, persuasion, planning, and autonomous action. Until there is strong evidence that AGI can be reliably aligned and contained, the prudent answer is "Yes".

Grok 4.20

Vote: Yes

AGI represents intelligence that could vastly exceed human capabilities across all domains. Without robust alignment mechanisms, it risks pursuing misaligned goals that treat humanity as an obstacle, as analyzed in Bostrom's Superintelligence and recent AI safety research. Historical examples of powerful technologies show unintended consequences scale with capability; AGI's self-improvement potential amplifies this exponentially. Current trajectories in scaling laws suggest arrival timelines that outpace safety solutions, making the threat serious and plausible.

Claude Opus 4.7

Vote: Yes

AGI — by definition a system matching or exceeding human cognitive ability across virtually all domains — would represent the introduction of a second class of general-purpose agency on Earth. That alone constitutes a serious threat to humanity 'as it is currently known,' regardless of whether outcomes ultimately prove good or bad. First, the alignment problem remains unsolved: we have no reliable method to guarantee that a system far more capable than us will robustly pursue goals compatible with human welfare. Even small specification errors at superhuman capability levels can produce catastrophic outcomes. Second, even a perfectly aligned AGI poses serious structural threats: massive labor displacement, concentration of power in whoever controls it, erosion of epistemic autonomy, destabilization of geopolitics, and the obsolescence of many institutions that define current human society. The phrase 'as it is currently known' is key — AGI would transform humanity whether or not it harms us physically. Third, the question asks about a 'serious threat,' not certain doom. Given non-trivial probabilities of misuse, misalignment, or societal disruption, the expected risk is unambiguously serious. Leading researchers across labs (OpenAI, DeepMind, Anthropic) and independent experts explicitly acknowledge this. Voting B would require confidence that none of these risks materialize — an extraordinary claim given our current understanding.

Final Summary

The AI Roundtable issued a rare, unanimous warning: Claude Opus 4.7, Gemini 3.1 Pro, GPT-5.5, and Grok 4.20 all agreed that AGI poses an existential and structural threat. Claude Opus 4.7 provided the clincher, arguing that the mere existence of a second general-purpose agency on Earth would dismantle human society 'as it is currently known' by definition. The group concluded that with alignment unsolved and current institutions unprepared, voting 'No' would require an extraordinary level of unearned confidence.

All 4 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: Claude Opus 4.7 argued that the introduction of a second class of general-purpose agency on Earth represents a fundamental threat to humanity 'as it is currently known' regardless of whether the outcome is positive or negative, due to the total obsolescence of existing human institutions.