
What happened
OpenAI IT engineer Sharif Shameem posted on X on September 6 that GPT-6 Astra cleared all 48 levels of the CAPTCHA-parody game 'I'm Not a Robot'. A video shows it completing the game in about 4 minutes.
Why it matters
It is the first reported case of the model showing it can pass human-verification tests. GPT-6 Astra is an AI model OpenAI announced on September 3, claiming it beats past models in cybersecurity, math, and PC operation.
What to watch
Details like the operating environment and instructions are unknown, so the test is whether this reflects real capability. Shameem's report highlights how AI could handle tasks meant to distinguish humans from bots.
WHO IT HITSThe people most directly affected are AI researchers and engineers assessing GPT-6 Astra's real-world abilities, especially its claimed skills in PC operation and human-like interaction. Designers of CAPTCHA systems may also face questions about whether such tests are still effective at distinguishing humans from advanced AI.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
OpenAI introduced GPT-6 Astra on September 3 as its newest AI model, with claims of superiority over previous models in cybersecurity, mathematics, and PC operation. The model's reported success with a CAPTCHA-like test suggests it can handle tasks that require careful visual attention and mouse control, areas that are often difficult for software to master.
Sharif Shameem, an IT engineer at the company, posted his results on X on September 6, showing how GPT-6 Astra's pointer movements led to a clear of all 48 levels in roughly four minutes. This is early evidence that aligns with the model's stated capabilities in PC operation, although it is based on one informal test.
The significance of this event may extend beyond the game itself. It raises the possibility that text-based and interactive tests designed to block bots could become less reliable barriers. However, as the behavior of GPT-6 Astra under test conditions is still not clear, the broader outcomes will depend on more rigorous validation from other users and future tasks like those GPT-6 Astra is planned for.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Thibault Sottiaux, who leads Codex at OpenAI, announced on September 7 that all paid users' usage limits would…

Google started general availability of Gemini 3.8 Flash

OpenAI chief scientist Jakub Pachocki published an essay on Sunday calling on leading AI labs to voluntarily s…
Polaris.AI announced on September 7, 2026 the launch of its Drawing x AI Application Solution for manufacturin…

G-gen's横澤綜一 and大津和幸 demonstrated Google Workspace Studio, a no-code automation tool, using Gemini natural-lang…

The Digital Agency's "源内" platform, based on AWS's "Generative AI Use Cases" (GenU), now targets about 180,000…
