Within the span of a few days this week, two completely separate events unfolded in Washington DC that could easily have happened in parallel universes. On one side of town, Silicon Valley executives, Trump administration officials and members of Congress gathered to celebrate AI's promise, touting productivity gains and American competitiveness. A day or two later, a handful of senators and labour leaders met in a hotel ballroom to strategise about how to fight back as the same technology threatens to remake the workforce. Bloomberg's account of this split captures something real: the politics of AI are no longer simply pro-innovation versus cautious. They are now shaped by two genuinely incompatible stories about who benefits.
Senator Mark Warner offered one of the week's more striking anecdotes. A venture capitalist, he said, recently told him he was writing software investments down to zero because of advances made by Anthropic's Claude. A major law firm told Warner it had stopped hiring first-year associates because AI can now handle much of the work those junior lawyers once did. These are not hypothetical future risks; they are decisions being made now, by real businesses, about real people. Warner's proposed response is to tax the data centres powering the AI boom and channel that revenue towards workers displaced by it.
Meanwhile, the Washington Post this week published a detailed look at Elon Musk's race to scale Tesla's Optimus humanoid robot programme. The story paints a picture of an industry that isn't waiting for regulation to catch up. Musk is building what the Post describes as a "robot army", with ambitions to have Optimus working in Tesla factories at scale and eventually in homes. Several other tech firms are following the same playbook. The urgency is commercial, not philosophical: whoever deploys physical AI at scale first captures enormous productivity and cost advantages, creating strong incentives to move faster than policy can track.
The tension between these two tracks is not new, but it is sharpening. What makes this moment distinct is that the labour concerns are no longer abstract. Employee fears about AI-driven job loss have jumped from 28 percent in 2024 to 40 percent in 2026, according to Mercer's Global Talent Trends research. Anthropic's own economic index, published earlier this month, found that AI is rapidly changing how work gets done even if it has not yet caused large-scale job losses at the aggregate level. That "not yet" is carrying a lot of weight right now.
The policy debate has also shifted onto trickier ground. The White House wants to codify a national AI framework this year, including child safety rules, data-centre permitting standards and intellectual-property protections. But the framework's primary political goal seems to be pre-empting state-level regulation, not addressing labour displacement. That framing suits the tech industry, which has consistently argued that a patchwork of state laws would cripple innovation and hand advantages to China. Whether federal pre-emption actually serves workers is a different question entirely, and it is the one being loudly asked in the hotel ballrooms.
What is not yet clear is whether the Warner-style "tax the data centres" approach can survive contact with a Congress that is broadly aligned with the tech industry's interests. The more interesting dynamic to watch may be whether labour's concerns start shifting the political calculus in swing states where manufacturing displacement is already visible. AI has been a bipartisan enthusiasm so far. That bipartisan consensus is looking a little less stable than it did a few months ago.