Anthropic sounds the alarm as the AI bubble takes off
Anthropic, creator of Claude, warned about unchecked AI development in a blog post, suggesting a global freeze could be beneficial if coordination were possible. The company also restricted its advanced AI model, Mythos, to a private group called Project Glasswing to prepare defenses against potential misuse by adversarial actors like China and others.
Anthropic, the U.S.-based developer of the AI model Claude, published a blog post titled *When AI Builds Itself*, highlighting concerns about AI systems increasingly writing their own code and proposing experiments, reducing human oversight. While the post acknowledged that slowing AI development might be beneficial, it noted the lack of a feasible global coordination mechanism, especially with countries like China, Russia, Iran, and North Korea involved. The company recently completed Mythos, an AI model described as 'strikingly capable' at computer security tasks, including hacking, but chose not to release it publicly. Instead, Mythos was shared with a select group of insiders under Project Glasswing, a private initiative aimed at safeguarding democracies from AI threats. The project includes 150 organizations, with a small number of Australian firms invited to participate, though selection criteria remain undisclosed. Anthropic CEO Dario Amodei framed Project Glasswing as a defensive measure to counter AI advancements by authoritarian regimes, drawing parallels to historical efforts like the 1944 Bretton Woods conference and the development of Colossus, the world’s first programmable computer, which aided Allied code-breaking during World War II. The initiative reflects a shift from government-led AI development to private-sector-led efforts, with Anthropic now positioning itself as a leader in AI, surpassing competitors like OpenAI. The announcement of Project Glasswing and Mythos’ restricted access came on April 7, coinciding with renewed stock market speculation in AI, following a surge in AI-related share prices on March 30. Anthropic’s actions underscore growing concerns about AI’s rapid advancement and the need for controlled, defensive strategies to mitigate risks, particularly in an era where AI could reshape global power dynamics.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.