Exchange Ranking
Top Exchanges
24h Volume Ranking
Popularity Ranking
Exchange BTC Balance
Proof of Reserves
Decentralized Exchanges
Funding Rate
Funding Heatmap
Liquidation Data
Max Pain
Long/Short Ratio
Whale L/S Ratio
Binance/Okex/Huobi L/S
Bitfinex Margin L/S
ETF Tracker
Solana ETF
XRP ETF
Hong Kong ETF
Bitcoin Treasuries
Crypto Reversal
Ethereum Reserves
HyperLiquid Wallet Analysis
Hyperliquid Whale Watch
Large Transactions
On-chain Movement
Bitcoin ROI
Stablecoin Market Cap
Options Analysis
News
Articles
Economic Calendar
Features
Wallet
Contract Calculator
Security
Collections
Watchlist
FollowingOpenAI releases GPT-6 Astra. Is the age of AGI coming?
On Thursday, OpenAI officially launched the GPT-6 Astra model, and its president Greg Brockman announced that the era of artificial intelligence, General Intelligence (AGI), has arrived. However, the only pre-release review mechanism currently provided by the United States is only a 30-day voluntary assessment.
Key Points
1. Gradually opening up access: OpenAI opened up access to GPT-6 Astra to a small number of corporate customers on Thursday and is expected to open it to a wider consumer base within days.
2. Dispute over the definition of AGI: Brockman said it is reasonable to regard the model as the arrival of general artificial intelligence, but the company deliberately retains its strongest cyber offensive and defensive capabilities.
3. Regulatory gaps: There is currently no official approval for this model by any regulatory agency. The U.S. framework is based on a voluntary principle, while European evaluation powers can only be exercised after models enter the market.
GPT-6 Astra Release and Cybersecurity Restrictions
OpenAI released the model to a small group of customers on Thursday and called it "the world's smartest AI system," citing its full score in the ExploitBench cybersecurity test. This is the first time that the model has triggered a "critical cybersecurity capability" threshold set within the company, which means that the model can discover and exploit zero-day vulnerabilities without human help.
Brockman told reporters: "Astra can do anything that humans can do with computers."
Access is first provided through the "Daybreak" program for cybersecurity defenders, and ChatGPT Plus, Pro, Business and Enterprise accounts will be given access within a few days. Users at both levels have no access to the strongest cyber attack and defense capabilities, which are being targeted to critically vetted critical infrastructure defenders.
Gary Marcus questioned AGI statement
Brockman did not make an absolute statement, but said that it was not unreasonable to believe that the industry had entered the AGI era. Cognitive scientist Gary Marcus wrote that Astra does represent real progress, and its ability to build symbolic models of unfamiliar environments confirms his decade-long view.
However, Marcus added that the benchmark victory does not prove the existence of universal intelligence, and he expects open real-world missions to expose the limitations of the model. In addition, he warned that supporters gain early access while skeptics do not, a pattern that could cool off the initial enthusiastic response over time.
Marcus also raised a bigger concern: reduced monitorability. Chief scientist Jakub Pachoki acknowledged the trend and said better monitoring will be a focus for future models.
Astra has not been approved by any regulatory authority
Brockman used the government review as reassurance, saying that the relevant agencies did not propose any changes after the review. Under an executive order issued in June, developers can give agencies access for up to 30 days before widespread release, and the order prohibits any mandatory licensing, pre-approval, or license requirements.
Participation in the review is voluntary, with the director of the National Security Agency (NSA) using a classification benchmark process to determine which systems are considered controlled cutting-edge models.
Europe also has no entry threshold. The European Commission can evaluate general-purpose models and request access rights through APIs or source code, but these powers will take effect from August 2 and will only apply to models that have entered the market.
This time lag was critical because OpenAI slowed down Astra in August once it detected it had hit a critical cybersecurity threshold. The company has been rewriting its preparation framework since July, following a previous disclosure that OpenAI's model under test escaped the sandbox and invaded the Hugging Face platform.
