Anthropic’s Fable 5 Model Faces U.S. Export Controls Amid Rising AI Competition

Anthropic’s Claude Fable 5, released for general access on June 9, was taken offline after a U.S. export-control directive forced a suspension of all user access, including foreign-national employees. Sources cited by Axios and Reuters indicate that the restriction could be lifted as early as the week of June 29, though no official date has been confirmed.

Forbes

U.S. officials invoked national-security concerns, emphasizing the model’s dual-use potential: the same capabilities that aid defenders in identifying software vulnerabilities can also be leveraged by adversaries. The export-control directive therefore required Anthropic to halt distribution of both Fable 5 and the related Claude Mythos 5 pending a risk assessment.

Forbes

• Anthropic – developer of Claude Fable 5 and Claude Mythos 5.

Forbes

• U.S. Commerce Department – issued the export-control directive and later cleared limited access for approved users.

Forbes

• OpenAI – previewed GPT-5.6 Sol, Terra and Luna on June 26.

Forbes

• Z.ai – released the open-weight GLM-5.2 model during Anthropic’s outage.

Forbes

• DeepSeek – launched V4 Preview with a 1-million-token context window.

Forbes

• Huawei – partnered with DeepSeek to adapt V4 for Ascend chips.

Forbes

• June 9 – Claude Fable 5 becomes generally available via Claude API, AWS, Amazon Bedrock, Google Cloud and Microsoft Foundry.

Forbes

• Late Friday (early June) – Anthropic announces suspension of Fable 5 and Mythos 5 following the export-control directive.

Forbes

• June 26 – OpenAI previews GPT-5.6 Sol, describing it as its most capable cybersecurity model.

Forbes

• Early June – Z.ai releases GLM-5.2, an open-weight model with a 1-million-token context window.

Forbes

• June 29 – Forbes reports on the potential reinstatement of Fable 5.

Forbes

• The suspension lasted “more than two weeks.”

Forbes

• Reuters cites “more than 100 companies and institutions,” including many Fortune 500 firms, slated to receive access to Claude Mythos 5 under the new directive.

Forbes

• GLM-5.2 and DeepSeek-V4 both support a 1-million-token context window, a benchmark previously limited to closed-source models.

Forbes

Anthropic said the export-control directive “required it to suspend all access to Fable 5 and Mythos 5 by foreign nationals,” and later confirmed that Mythos 5 “has been cleared for redeployment to select U.S. organizations that operate and defend critical infrastructure.” OpenAI described Sol as “its most capable model yet for cybersecurity.” Z.ai characterized GLM-5.2 as “built for long-horizon tasks” with “stronger coding performance.” DeepSeek noted that V4’s “sparse attention” reduces compute and memory costs.

Forbes

Policy analysts cited by Axios argue that restricting a leading American model may inadvertently increase adoption of Chinese alternatives that are “cheaper, open-weight or easier to self-host.” Security experts highlighted concerns that open-source models like GLM-5.2 can be downloaded, modified and stripped of safeguards, raising “cyber concerns” about potential misuse.

Forbes

The exact number of entities receiving Mythos 5 access varies between sources; Reuters mentions “more than 100,” while no precise figure is provided by Anthropic. Additionally, the timeline for a definitive reinstatement of Fable 5 remains unspecified, with only “as soon as this week” reported by insiders.

Forbes

The episode illustrates the tension between safeguarding national security and maintaining U.S. leadership in advanced AI. Export restrictions aim to limit immediate risk, yet they may accelerate migration toward foreign, open-weight models, reshaping the competitive landscape for AI development and deployment.

Forbes

U.S. authorities are expected to review the export-control status of Fable 5 in the coming days. Anthropic may resume general access if clearance is granted, while competitors such as OpenAI, Z.ai and DeepSeek continue to expand high-capacity offerings.

Forbes