Anthropic Halts Launch of Fable and Mythos Amid New US AI Safety Order

In a move that surprised few in the tech world but still raised eyebrows, Anthropic announced on Thursday that it is putting the brakes on two of its most‑anticipated AI products – the conversational model codenamed Fable and the image‑generation system dubbed Mythos. The decision comes directly after a fresh directive from the White House’s Office of Science and Technology Policy (OSTP), which asked all leading AI labs to pause the deployment of any system that could out‑perform today’s most capable models, like GPT‑4.

“We’re fully committed to responsible innovation,” said Anthropic CEO Dario Amodei in a brief statement. “If the government says we need a deeper look at the risks, we’ll take that seriously and work closely with regulators.” He added that the company is already conducting internal safety assessments, but that the timeline for releasing Fable and Mythos is now uncertain.

The two models were supposed to be Anthropic’s next big step forward. Fable, a next‑gen chatbot, promised richer, more nuanced dialogues that could, in theory, handle complex customer‑service scenarios or even tutoring. Mythos, on the other hand, was geared toward creating high‑fidelity images from textual prompts, a capability that could rival offerings from OpenAI’s DALL‑E or Google’s Imagen. Both were slated for a limited beta later this year, according to earlier internal roadmaps.

But the OSTP memo—issued under the broader “AI Bill of Rights” initiative—doesn’t just ask for a polite check‑in. It explicitly demands that any model deemed “more capable than GPT‑4” undergo a rigorous, third‑party risk assessment before being released. The government’s language is clear: safety first, even if it means slowing down the race for bigger, flashier AI.

Anthropic isn’t the only company feeling the pinch. OpenAI, which recently unveiled GPT‑4 Turbo, has also said it will halt certain upgrades until it can demonstrate that the new features meet the new safety thresholds. Google’s DeepMind, Microsoft, and a handful of other labs have echoed similar sentiments, indicating that the directive may reshape the entire industry’s rollout schedule.

For now, Anthropic says it will keep its existing models—Claude 2 and the older image generators—available to customers, while the new research prototypes stay under wraps. “We’ll continue to iterate, learn, and share what we discover,” Amodei added, hinting that the pause could actually be a chance to build stronger safeguards rather than just a bureaucratic hurdle.

Analysts are watching closely. Some argue that this governmental nudge could prevent a hasty, potentially dangerous sprint toward ever‑more powerful AI, while others worry it might put U.S. firms at a competitive disadvantage compared to overseas rivals not subject to the same rules. Only time will tell how the balance between innovation and caution will play out, but for Anthropic and its peers, the message is crystal clear: safety isn’t optional—it’s now a prerequisite for any next‑generation launch.