Anthropic's Claude 5 Restrictions May Be Lifted This Week
The U.S. government is expected to lift access restrictions on Anthropic's top-tier model, Claude 5, as early as this week. According to a report by Axios on June 30, the Trump administration has moved toward restoring access to Claude 5.
On June 26, Commerce Secretary Gina Raimondo sent a letter to Anthropic allowing limited redistribution of its cybersecurity-focused model, Mythos 5, to about 100 agencies responsible for operating and defending critical U.S. infrastructure. However, as of June 30, access to Claude 5 remains blocked for the general public, pending final approval from the Pentagon and NSA.
Both Anthropic and OpenAI are jointly urging the administration to establish a transparent and systematic legal review process, as outlined in President Trump's executive order from June 2, moving away from the current case-by-case review method.
Google's Gemini 3.5 Pro Launch Delayed Until July
As of June 30, Gemini 3.5 Pro remains in a limited preview phase for select Vertex AI enterprise customers. Google aims for an official launch in July, though no specific date has been announced. Sundar Pichai, CEO of Google, had previously promised during the May I/O event that the model would be available the following month, but that commitment has not been fulfilled.
Experts suggest that the GPT-5.6 model achieved a score of 96.7% on the U.S. government's cybersecurity benchmark, surpassing the control threshold, while Google's latest production model did not meet that standard and was thus excluded from government review.
Colorado AI Law Significantly Rolled Back Before Implementation
Colorado Governor Jared Polis signed SB 189 on May 14, postponing the original June 30 effective date of the state's AI law to January 1 of next year and significantly relaxing key regulatory requirements. The amendments removed existing obligations for AI providers, such as the duty to prevent algorithmic discrimination, maintain risk management programs, and conduct impact assessments, shifting to a narrower framework focused on the disclosure of automated decision-making technology (ADMT) usage. Once regarded as the first comprehensive AI regulatory law in the U.S., Colorado's AI law has effectively retreated from its original provisions amid federal pressure and corporate pushback.
Austria Urges EU to Attract Anthropic Amid U.S. Restrictions
Austria's Minister for Digitalization, Alexander Schallenberg, cited the U.S. restrictions on the Claude model as a direct backdrop for sending an official letter to European Commission Vice President Henna Virkkunen, requesting a review of establishing Anthropic's European Union base. No official response from the EU has been received yet.
Anthropic Claims Alibaba AI Model Faced Large-Scale Distillation Attack
Anthropic has alleged that approximately 25,000 fake accounts systematically analyzed the reasoning, coding, and complex task performance capabilities of its Claude model through over 28 million interactions. Alibaba has not publicly responded to these claims.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.