OpenAI to Achieve AGI by End of 2026, Altman Predicts

OpenAI CEO Sam Altman stated that the company will achieve AGI (artificial general intelligence) by the end of 2026, according to TIME reporting. However, this prediction comes with an important caveat: it assumes acceptance of what 'AGI' means based on Altman's own interpretation of the definition.

AGI refers to the concept of an AI system that possesses intellectual capabilities equal to or exceeding those of humans and is capable of addressing challenges across all domains. Nevertheless, there remains ongoing debate among researchers and companies about 'what level of achievement qualifies as AGI,' with no universally agreed-upon definition. The fact that Altman's statement is qualified as being 'by his definition' directly reflects the difficulty of this debate.

The report also includes comments from Jakub Pachocki, chief scientist at OpenAI. He stated that Astra, a newly developed model that OpenAI is creating, is already functioning as an 'automated research intern.' This means it serves the role of an AI agent assisting researchers. Furthermore, Altman positioned Astra as 'the first model to become capable of actually inventing something genuinely new in a meaningful way.'

The attention this statement receives stems from industry-wide interest in AI's capacity for 'autonomous discovery and invention.' Until now, AI systems have learned from vast datasets and generated responses by combining existing knowledge. By contrast, reaching a stage where AI can 'generate novel insights that no one yet knows'—as human researchers do—represents a qualitative turning point in AI capabilities. Altman's assessment of Astra can be interpreted as demonstrating confidence in reaching that inflection point.

However, it is important to note that whether OpenAI declares it has achieved AGI depends not only on the technological reality but also on the definition question. Within the industry, the term 'AGI' is used with different meanings by different companies and researchers, and Altman's remarks must be understood within that context. While specifying the concrete timeframe of 'end of 2026' is clear, the matter of who evaluates and how they assess that achievement remains a separate issue.

The focus going forward lies on when and at what stage Astra's actual capabilities will be publicly unveiled and verified, and in what form the claim of 'inventing something new' will be demonstrated. Even as definitional discussions surrounding AGI continue, the fact that OpenAI is accelerating development with 2026 as a key milestone can be seen as likely to influence the competitive landscape of the entire industry.