YouTube video by Deep Press Analysis

Members of the House Committee on Foreign Affairs urged passage of the Chip Security Act, citing recent public disclosures and DOJ charges they say show roughly $2.5 billion in illicit exports of advanced AI chips to China and arguing the bill would tighten export safeguards without mandating specific technical solutions.

With a significant shift in U.S. government policy toward frontier artificial intelligence deployment, limited access has been restored to Anthropic's advanced Mythos 5 model, signaling a more targeted regulatory strategy than a blanket ban.  Following a suspension of the model earlier this month due to national security concerns, U.S. authorities have now authorized its release to a carefully vetted group of organizations, including major Fortune 500 companies, which have been carefully vetted.  Washington has emphasized the importance of balancing artificial intelligence innovation with national security safeguards, as increasingly capable foundation models are subject to increased scrutiny over their potential misuse by foreign military and intelligence entities.  Additionally, the move is a useful illustration of a growing trend in which governments are increasingly influencing the deployment of cutting-edge AI systems and in which access to those systems is increasingly linked to trust, security compliance, and controlled distribution rather than unrestricted public access.  Regulatory discussions prompted by the U.S. government's export control order issued on June 12, which required Anthropic to suspend access to both Mythos 5 and its companion model, Fable 5, while officials assessed the possible national security implications of releasing frontier artificial intelligence capabilities, led to the latest authorization.  As the administration noted, it was concerned that highly capable generative AI models could be exploited by military or intelligence agencies linked to China, Russia, and other countries considered strategic risks. In light of this, Anthropic sought to strengthen compliance measures with the U.S. authorities, ultimately obtaining approval from the Secretary of Commerce Howard Lutnick to reactivate Mythos 5 to a limited network of vetted partners.  However, Fable 5 remains subject to export restrictions while regulatory assessments are being completed. There has also been a broader shift in policy, as OpenAI announced it had postponed the full public rollout of GPT-5.6 at the request of U.S. officials, limiting early access to a small number of pre-approved organizations whose identities were disclosed to the government in response to the change.  Together, these developments demonstrate the growing regulatory framework for the deployment of frontier AI models, in which access to these models is increasingly restricted, government oversight is continuous, and available models are available to a narrower audience rather than being made available widely to the public.  While the government has reversed the partial policy, its selective approval process continues to polarize discussion over the need for transparency and competitive fairness as frontier AI models are deployed. As a consequence of the lack of clearly defined eligibility criteria, federal agencies have accumulated considerable discretion, leaving companies outside the approved ecosystem with little insight into the decisions made regarding access.  As a legislative counsel for the Foundation for Individual Rights and Expression, John Coleman has questioned the opaque vetting framework, arguing that a lack of transparency in participant selection raises broader concerns about accountability and the consistency of regulatory authority application.  Achieving the same objective, Commerce Secretary Howard Lutnick confirmed that organizations on the approved list of trusted organizations, as well as their employees, including non-U.S. citizens, as well as Anthropic's own international workforce, will be exempt from requiring individual export licenses to access Mythos 5.  Licensing requirements, however, will remain in force for organizations outside of the government's trusted network. A number of the approved entities have been participating in Anthropic's Project Glasswing initiative, a collaborative effort between approximately 100 established technology companies and research institutions. It is also being discussed whether or not Fable 5 will be authorized in the future, although no implementation dates have been disclosed. Increasing national security concerns increasingly influence commercial deployment strategies, which is reflected in the evolving regulatory framework which reflects a broader shift in how advanced artificial intelligence capabilities are governed. Although Fable 5 and Mythos 5 are based on the same underlying foundation model, the latter has been designed to be widely available with fewer deployment restrictions, making its continued suspension a noteworthy distinction in the government's risk assessment.  A number of regulatory frictions have also resulted from Anthropic's refusal to support the use of its AI models for domestic surveillance and fully autonomous weapons systems. This stance exacerbated frictions between Anthropic and Washington. Additionally, both Anthropic and OpenAI continue to pursue public market ambitions while adjusting to the new compliance requirements introduced in President Donald Trump's executive order.  By establishing a voluntary framework, the U.S. government will have the opportunity to review frontier artificial intelligence models up to 30 days before they are released to trusted partners under this voluntary framework. Analysts point out that while the latest authorization provides a practical mechanism for controlled deployment in the near-term, it does not resolve the question of how advanced AI systems are able to be deployed at scale.  A former Commerce Department official and analyst at the Center for Strategic and International Studies, Ms. Koren warned that prolonged uncertainty surrounding broad model deployment could eventually erode the competitive advantage of U.S. AI developers. This could create opportunities for geopolitical rivals such as China to narrow their technological gap.  Advance AI models are progressively being returned under tightly controlled access, signaling that frontier artificial intelligence has entered a new era where technical capability alone is no longer the determining factor of deployment.  As governments refine oversight mechanisms for high-impact AI systems, developers, enterprises, and security teams must adjust to ever-evolving compliance requirements. Those considering integrating next-generation artificial intelligence need to closely monitor regulatory developments, export controls, and trusted access frameworks, as policy decisions are becoming an increasingly important aspect of AI adoption.

CIOs at organizations expanding their AI efforts won’t see any short-term benefits, but US AI dominance over China may bring innovation advantages over the long run.

Claude maker warns authoritarian regimes could set the rules unless Washington tightens chip and model controls

Full coverage of global sanctions regimes, including the United Nations, European Union, United Kingdom, United States, and many others.

Anthropic submits detailed recommendations for strengthening US export controls on advanced AI chips and model weights. We advocate for maintaining America's compute advantage, adjusting tiering syste...