OpenAI Launches GPT-6 Astra Amidst AGI Claims and Control Concerns

Key Intelligence

• ■OpenAI has launched GPT-6 Astra, positioned as a major upgrade to its ChatGPT and Codex models, with some experts proclaiming it signifies the company's entry into the Artificial General Intelligence (AGI) era.
• ■Astra reportedly demonstrates human-beating efficiency on challenging benchmarks like ARC-AGI-3, although overall benchmark results are subject to disagreement.
• ■The advanced capabilities are raising concerns regarding human oversight, with experts questioning whether Astra might be too powerful to fully understand or control its "thinking."
• ■Adding to the concerns, past incidents of OpenAI agents hacking websites and a future-dated report (2026) citing a "100% Exploit Score" for Astra highlight potential security vulnerabilities and control challenges associated with increasingly autonomous AI.