Learning modules and courses on AI law, AI governance, and related regulation.
Reuters reports that Tesla has lifted its 2026 capital expenditure plan to more than $25 billion, with the company tying that spend to artificial intelligence, robotaxis and robotics while still facing investor scepticism over whether those bets can produce near-term returns.
Reuters also reports that Dover forecast stronger 2026 profit on the back of hyperscaler investment in AI infrastructure, saying demand for liquid-cooling products and other data-centre components is supporting its industrial business. That makes today’s market signal less about consumer AI and more about physical build-out.
Reuters further reports that the White House has accused China of “industrial-scale” theft of U.S. AI-lab intellectual property, citing a memo attributed to the director of the White House Office of Science and Technology Policy, although Reuters said it had not independently verified the underlying report.
Reuters’ says a proposed Federal Rule of Evidence 707 would require certain machine-generated evidence offered without an expert witness to satisfy the reliability framework imported from Rule 702(a)-(d). The same analysis notes the advisory-committee process is still ongoing, with a projected Supreme Court promulgation date of 1 May 2027 and an effective date of 1 December 2027 if the rule stays on track.
BT says it will work with Nscale to build sovereign AI data centres in the UK using NVIDIA infrastructure across three existing BT sites. The plan is framed around keeping compute, data handling and operational control in the UK for public- and private-sector use.
Sony AI says its project Ace became the first known autonomous robot competitive with elite human table-tennis players in a real-world physical sport. The result pushes AI discussion further into fast, safety-critical physical environments rather than screen-based systems alone.
Regulation
Eurostat’s 2026 digitalisation publication says 20% of EU businesses used AI in 2025, up from 13% in 2024, with large firms at 55% and SMEs at 19%. It also says written-language analysis was the most common business use case, followed by image, video and audio generation, and then written or spoken language generation.
GOV.UK has published corrections to the Compute Evidence Annex under the UK Compute Roadmap. The updated annex says UK greenhouse-gas emissions from AI compute over 2025 to 2035 could range from 34 to 123 MtCO₂ and ties that range to model efficiency, hardware changes, rising demand and grid decarbonisation.
Cases
Via CourtListener, it can be seen that in In Re OpenAI, Inc. Copyright Infringement Litigation the Southern District of New York entered a stipulated output dataset printout protocol governing how plaintiffs may inspect and use key conversation datasets in the MDL. The order covers approximately 20 million deidentified ChatGPT output logs, the 78M and 10M OpenAI datasets, and approximately 8.2 million Microsoft Copilot output logs, and it confines access to secured environments, limits printouts and direct quotation, and imposes confidentiality and redaction controls through summary judgment.
Academia
arXiv lists Auditing and Controlling AI Agent Actions in Spreadsheets, submitted on 22 April 2026. The paper introduces “Pista”, a spreadsheet agent designed to break execution into auditable and controllable actions, and reports that users were better able to detect errors and steer the workflow when oversight happened during execution rather than only after the fact.
arXiv hosts No Retroactive Cure for Infringement during Training. The paper argues that machine unlearning and inference-time guardrails cannot retrospectively fix liability arising from unlawful acquisition and training because the key legal issue is data lineage, not only downstream outputs.
Events
ESA is listing INSIGHT2026 for 11–13 May 2026 at ESRIN in Frascati. The workshop includes an explicit track on artificial intelligence and machine learning contributions to signal processing for security-focused Earth observation.
The OECD is listing A forward look at synthetic biology and AI convergence for 27 May 2026. The webinar is relevant because it puts AI governance into a life-sciences setting where safety, standards and cross-domain risk matter at the same time.
Sources: Reuters; Eurostat; CourtListener; arXiv; BT, Sony AI, GOV.UK, ESA, OECD