OpenAI published a 13-page policy document titled "Industrial Policy for the Intelligence Age," calling for a public wealth fund distributing AI profits to citizens, a 32-hour workweek with full pay, taxes shifted away from labor toward capital and automated labor, and a New Deal-scale grid buildout. Sam Altman frames superintelligence as more disruptive than the Industrial Revolution. Critics — including some economists — are calling it cover for regulatory nihilism.
Anthropic announced its annualized revenue run rate has tripled from $9B (end of 2025) to $30B, and sealed a massive compute deal with Google and Broadcom for 3.5GW of TPU capacity starting in 2027 — on top of 1GW already online. The number of enterprise customers spending over $1M/year doubled to 1,000+ in just weeks. This is the most significant AI infrastructure commitment announced by a non-hyperscaler to date.
The Justice Department filed its appeal of Judge Rita Lin's injunction blocking the Trump administration's ban on federal use of Anthropic's Claude. The original ruling found the Pentagon's "supply chain risk" label — applied after Dario Amodei refused to allow Claude for autonomous weapons or domestic surveillance — constituted illegal political retaliation. The appeal escalates a constitutional showdown over government AI procurement as a coercion tool.
Researchers at Tufts University unveiled a neuro-symbolic AI system that cuts training energy to 1% of standard model requirements and slashes inference power to 5%, while achieving a 95% success rate on Tower of Hanoi tasks versus 34% for conventional models. The system combines neural networks with symbolic reasoning, more closely mirroring human logical decomposition. The work will be presented at ICRA in Vienna in May.
Anthropic's Claude.ai suffered a major outage on April 6-7, with elevated errors across web, desktop, mobile, voice mode, and login. This is the second notable service disruption in recent weeks. As Claude's enterprise footprint expands (see story #2), reliability incidents carry growing reputational and contractual weight.
Large health systems and independent therapists are rapidly adopting AI tools for session notes, record-keeping, and care management. Enthusiasm is real — administrative burden is crushing the sector — but a vocal contingent of clinicians and researchers is raising alarms about patient safety, data privacy, and the risk of AI embedding clinical bias into sensitive treatment contexts.
OpenAI opened applications for its Safety Fellowship, funding independent external researchers to pursue AI safety and alignment work from September 2026 through February 2027. The program is positioned as an olive branch to the academic safety community, though some researchers will note the timing — announced the same week as OpenAI's industrial policy push — is conspicuously self-serving.
