Details Emerge on Anthropic's Model Block Triggered by Amazon
The situation surrounding the export controls on Anthropic's PaLM 5 and Mitos 5 models is becoming clearer. According to the IT industry on June 15, Amazon, a major investor and board member of Anthropic, discovered a method to bypass the PaLM 5 model and directly informed Treasury Secretary Scott Bessen. This report from Amazon led to the issuance of export control guidelines by Secretary of Commerce Howard Rootnick.
David Sachs, co-chair of the White House Council of Advisors on Science and Technology, claimed on social media that the government requested Anthropic CEO Dario Amodei to either fix the bypass or take down the model, but Amodei refused. Sachs criticized Anthropic's public statement that the bypass was not serious, stating, "That is not the judgment of a trusted partner or the government. Anthropic itself defined Mitos as a 'cyber weapon' and requested regulation, so it is contradictory to claim the bypass is not serious."
In this context, it was revealed that Andrei Karpathy, an AI scientist at Anthropic, was unable to access his company's model because he is not a U.S. citizen, highlighting the absurdity of the situation. Additionally, Defense Secretary Pete Hekes stated on social media, "We expelled Anthropic from the building three months ago. Every day, it is proven to be the right decision."
As of June 14, both models remain offline, and Anthropic has not announced a recovery schedule. Sachs indicated that if Anthropic addresses the vulnerabilities, export controls could be lifted soon.
Canadian PM Highlights Risks of Overreliance on Specific AI Models
During a visit to Ireland on June 14, Canadian Prime Minister Mark Carney told reporters that the recent U.S. export control situation starkly illustrates the strategic vulnerabilities of excessive dependence on a small number of U.S. AI providers. Carney noted, "The current Mitos and PaLM situation is something that can happen anytime if there is an overreliance on specific models," emphasizing the need for AI sovereignty and diversification of suppliers. Given that the situation has blocked access for all foreigners, including nationals from allied countries like Canada, the UK, and Australia, it is likely to heighten crisis awareness among nations reliant on U.S. big tech AI.
Trump Abruptly Cancels AI Executive Order Signing
President Donald Trump unexpectedly canceled an event in the Oval Office to sign a new AI executive order. He cited concerns that the draft of the executive order could undermine the U.S. competitive advantage in AI against China. The regulatory demands from the finance and banking sectors regarding AI cybersecurity risks triggered by the Anthropic Mitos situation were reportedly the backdrop for this executive order.
Google AI Overview: Traffic Drops 33%, Publisher Links Increased
Google has implemented measures to increase traffic to publisher sites through its AI overview update. This move is seen as a response to an analysis indicating a 33% global decrease in search referrals. At Google Marketing Live, it was confirmed that Gemini is powering the entire advertising stack and that interactive ad formats are replacing static placements. The issue of AI-generated search answers encroaching on web traffic is affecting the revenue structure across the advertising ecosystem, which is also relevant to domestic media and portal industries.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.