OpenAI Model Cracks "I'm Not a Robot" Captcha Test

Astra Clears 48 Levels of "I'm Not a Robot" Game Shows Visual and Analytical Skills Rivaling Humans "Demand Surging; New Paid Sign-Ups May Be Capped"

International|
|
By Park Min-joomj@sedaily.com
||
The OpenAI logo. Reuters-Yonhap News - Seoul Economic Daily International News from South Korea
The OpenAI logo. Reuters-Yonhap News

OpenAI's artificial intelligence model GPT-6 Astra has passed a Captcha test designed to tell computers apart from humans.

OpenAI developer Sharif Shameem said on X, formerly Twitter, on the 7th that Astra "successfully solved 48 levels of the 'I'm Not a Robot' game."

A screen capture of the "I'm Not a Robot" game posted on X by OpenAI developer Sharif Shameem. - Seoul Economic Daily International News from South Korea
A screen capture of the "I'm Not a Robot" game posted on X by OpenAI developer Sharif Shameem.

The "I'm Not a Robot" game, created by U.S. programmer Neal Agarwal, is a form of Captcha — a completely automated public Turing test to tell computers and humans apart, named after the test devised by mathematician Alan Turing to determine whether a machine possesses intelligence. It consists of tasks that are easy for humans but difficult for machines to solve, such as identifying distorted letters of the alphabet or selecting every image containing a particular object from a set of nine photos. The tests were designed to block crawling bots that scrape websites and mass spam distribution.

Astra's ability to solve the game suggests the model has acquired spatial and visual reasoning capabilities such as image recognition, contextual reasoning including comprehension of instructions, and computer-use skills on a level comparable to humans. Astra is also the first OpenAI model to be assigned a "Critical" safety rating. That has heightened the need for new technology to distinguish AI from humans.

After AI model benchmarking firm Artificial Analysis revised the evaluation framework for its intelligence index (AAII) twice over four days, Astra emerged tied for first place with Anthropic's Fable 5.1. Artificial Analysis said it had "brought forward some features of an upcoming evaluation framework in line with the pace of advances in frontier AI technology."

Meanwhile, Thibault Sottiaux, a key executive who led ChatGPT and Codex development at OpenAI, said on the 9th that demand for Astra is unprecedented. "We are doing everything we can to keep up with this demand, but we have never seen anything like this, even during periods of very rapid growth in the past," he said, adding that "if this continues, we may have to temporarily pause new Pro plan subscriptions for a while."

Original reporting by Park Min-joo for Seoul Economic Daily.

AI-translated from Korean. Quotes from foreign sources are based on Korean-language reports and may not reflect exact original wording.

Watch · Seoul Economic Daily

More →
3:12

AI KEY

Preview
Korean Corporate Intelligence HubKOSPI · KOSDAQ · 12 sectors

A live, cap-weighted view of every KOSPI and KOSDAQ sector, with same-day Korean reporting distilled by company — built for foreign investors, correspondents and analysts who need to scan Korea before the next session.

Korea Company Atlas

Preview
Market Ontology · The Feedback LoopKFTC 2025 · 92 groups · 121,954 articles

An English ontology of the Korean market — how companies, the media, the government and the National Assembly move each other in a loop. Korea's named controlling persons and designated business groups are a mechanism, not a risk to be priced blind.

SIGNAL

Now live
English Edition · Capital MarketsM&A · IPO · PE · Fund Flows

SIGNAL English Edition is live — Korea's deal desk reporting in English. M&A, IPOs, private equity and fund flows, covered daily for global institutional investors. Browse free; subscriber-only scoops at the 50% intro rate.