Explore our reporting on how AI and robotics change work, business and the world.
Anthropic says Claude helped reconstruct how several NIST and ANSI curve seeds were generated. Its code matches published values, while some base points and selection choices remain unexplained.
Six METR staff have employee-like access for risk assessments at Anthropic. The evaluator can publish findings, while Anthropic retains control over access and specified redactions.
Microsoft reports MAI-Base-2's held-out text results and fitted training-efficiency estimates. The base model is still in post-training, with no public checkpoint to test.
Google traced one peptide candidate through sequence, structure and genomic checks with Antigravity. Its predicted activity still awaits lab results.
Agility named Nyobolt for Digit 5's planned fast-charge battery system. Its run, charge and 20-hour daily-work figures are preliminary company claims, with early access expected in 2027.
The hosted Voice Agent API now uses Grok Voice 3.0. An independent simulated support score rises, while first-audio time increases and the listed voice price has a separate text charge.
CERN and Mistral signed a statement of intent for AI in physics. Existing CMS reconstruction work predates the proposed foundation-model partnership.
NVIDIA introduced early-access cuPotential and detailed partner work across chip design, manufacturing simulation and inspection. Public libraries and partner plans have different access paths.
ASML reported €11.7 billion in third-quarter sales and said customers' AI investment plans are lifting equipment forecasts. Its projected supply gap for 2027 and 2028 remains a management assessment, while interest in 2029–30 orders has yet to become orders.
Ai2's HiRO-ACEv2 adds wind and pressure to 3 km simulated rainfall. Its reported extreme-event matches are to X-SHiELD, with limits for forecasting and changed climates.
QuEra reports 695 on-target recoveries in 700 testbed trials of a deterministic laser-relocking script developed with Claude. Its separate tuning workflow still uses AI.
NVIDIA has committed compute access valued at $250 million over five years for UK agent safety research. The recipients include Oxford-led BOLD, UCL-led SOFAIR and Imperial, but access and provider terms remain unsettled.
ASE says its PowerSiP expansion lets designers place voltage-regulator modules beneath AI processors. Its claimed 50% or greater reduction concerns parasitic routing losses; the public pages give no test method or system-level measurement.
DeepMind and Singapore's GovTech describe a two-stage phone-based scam detector: quick message screening followed by Gemma conversation analysis. Its rollout and results are unknown.
In ISE's 2026 survey, 92% of UK student employers suspected or confirmed misconduct. The public findings show where AI use is usually permitted and what the figure cannot prove.
Liquid AI's new LFM2.5-24B-A2B has downloadable checkpoints for local and GPU serving. Its license and separate edge and server benchmarks matter when evaluating it for tool-using agents.
Anthropic's internal Scion system moves source-control history, commits and merges to a service while agent and CI clients fetch only the code they need.
Anthropic ran a model across 15 TV boxes and tested replication in a simplified emulator. The two results do not establish a rogue AI worm in the wild.
