White House lifts export control on advanced AI model from Anthropic

In a significant move, the U.S. government has lifted export restrictions on one of Anthropic's leading AI models, as announced by Commerce Secretary Howard Lutnick on Tuesday evening. This decision follows a thorough collaboration between Anthropic and government officials to evaluate the model, dubbed Fable 5, ensuring it meets safety standards while reinforcing America's position in the AI landscape. Fable 5 is a refined version of Anthropic's highly advanced AI model, Mythos, designed with additional safeguards for public application. Despite its capabilities, Mythos remains accessible only to a limited group of approved partners. Earlier this month, Anthropic was compelled to cease access to both Mythos and Fable following the implementation of an export ban by the government, which impacted all foreign nationals, including Anthropic employees. The ban was triggered by a report from a trusted partner, identified by CNN as Amazon, which uncovered a potential workaround, or 'jailbreak,' that could bypass Fable’s safety measures to enable Mythos-like functions. Anthropic acknowledged that these jailbreak methods were relatively straightforward and noted that similar vulnerabilities exist in other publicly available models. Following a series of discussions, the Commerce Department has now permitted Anthropic to release its potent Mythos AI model to specific government-sanctioned entities. These talks aimed to mitigate undisclosed risks tied to the model's features. The advanced capabilities of Mythos have raised alarms within both the U.S. government and financial markets, with experts warning that it could exploit cybersecurity weaknesses at an alarming rate. Initially, Anthropic had restricted Mythos's release to a select group of essential partners to safeguard critical software. This recent export ban and the negotiations that ensued highlight the ongoing challenges in establishing a cohesive regulatory framework for AI technology, especially as the U.S. seeks to maintain a competitive edge against nations like China. Additionally, the White House has urged OpenAI to restrict access to its forthcoming GPT 5.6 model to a small, government-approved circle due to its advanced capabilities. OpenAI expressed concerns that such government-imposed access should not become the standard practice moving forward.