Anthropic, the artificial intelligence startup founded by former OpenAI researchers, has announced that the United States government has lifted an export ban on its most advanced AI models. The development, confirmed by company sources late Tuesday, marks a significant shift in US export policy for cutting-edge AI technologies.
The ban, originally imposed under the Biden administration's Executive Order on AI Safety, restricted the export of frontier models to certain foreign entities over national security concerns. Anthropic declined to specify which countries were previously blocked but confirmed that the restrictions have now been removed for its latest models, including Claude 3.5 Opus.
The news has sent ripples through the UK tech sector, where firms are urging the British government to negotiate similar access. UK AI companies, including DeepMind and Graphcore, have argued that export controls risk hampering innovation and creating a two-tier global AI ecosystem.
"This is a moment of algorithmic apartheid," said Dr. Priya Patel, Director of the UK AI Safety Institute. "If the US unbundles its frontier models only for itself, we risk a digital iron curtain that slows progress on global challenges like climate modelling and drug discovery."
The US Commerce Department's Bureau of Industry and Security has not yet commented on the reported lifting of the ban. However, industry insiders suggest the move reflects a shift in strategy from blanket restrictions to targeted oversight.
For Anthropic, the policy change could accelerate its international expansion. The company has been positioning itself as a more safety-conscious alternative to rivals like OpenAI and Google DeepMind. Its constitutional AI approach, which embeds ethical guidelines directly into model training, has won favour with regulators.
But the lack of clarity on exactly which models are affected and under what conditions has caused confusion. Some experts worry this could be a precursor to a more fragmented AI landscape where nations hoard their best models.
"We are sleepwalking into a fragmentation of the global AI commons," warned Julian Vane, Technology and Innovation Lead. "The user experience of society will be one of haves and have-nots. Quantum computing and AI should be the shared infrastructure of our future, not weapons in a cold war."
The UK government has remained tight-lipped on whether it will seek a similar bilateral agreement with Washington. But sources inside Number 10 indicate that the Prime Minister raised the issue during his recent meeting with US Vice President Kamala Harris at the AI Safety Summit in San Francisco.
Meanwhile, UK tech firms are growing restless. A coalition of British AI companies has drafted an open letter to the Secretary of State for Science, Innovation and Technology, demanding a level playing field.
"We cannot have a scenario where British hospitals and universities are denied access to the best cancer detection AI while American institutions benefit," said Sarah Bennett, CEO of London-based health AI startup SynthRx.
The timing of the US decision is curious. It comes just weeks before the next round of international AI negotiations at the G7 Digital and Tech Ministers' meeting in Tokyo. Critics argue that the unilateral move undermines multilateral efforts to create shared AI governance frameworks.
For now, the ball is in the UK's court. Whether it can negotiate similar access will depend on the strength of the transatlantic relationship and its ability to offer reciprocal benefits to US firms.
What remains clear is that the geopolitics of AI are accelerating faster than most governments can adapt. The question is not just who gets the algorithms, but who gets to shape the reality they will build.
