The latest verified reporting and analysis across ai.
Energy North has withdrawn the June 2028 first-power date for its proposed off-grid AI campus in Australia’s Northern Territory. The project is not cancelled, but its 1-gigawatt computing plan now depends on an environmental review whose timetable has been suspended while regulators set the rules.
Read story
OpenAI, Google, Microsoft and Anthropic are converging on one machine-readable signal for generative-AI training preferences. The proposal could simplify publisher controls, but it is not yet an IETF standard, does not block crawling or grounding by itself, and creates no licensing or payment right.
Read story
A WIPO Secretariat report gives patent offices a menu of choices for proving human contribution, handling AI-generated disclosures and deciding when routine access to AI makes an invention obvious. The options now go before member states; they are not new global rules.
Read story
The trillion-parameter model can now run on private infrastructure and reproduce real software flaws before patching them. Mistral says its safeguards distinguish defensive from malicious work, but downloadable weights cannot be recalled or centrally monitored.
Read story
The public Lean repository could turn mathematics’ ‘enormous theorem’ into a machine-checkable dependency graph. But the release is still self-assessed, and its strongest claim now depends on independent reproduction.
Read story
The AI system can now propose and prove properties for C software and Verilog hardware. Machine-checked proofs raise assurance, but only for the specifications, models and assumptions engineers actually review.
Read story
A comparative study prepared for WIPO’s copyright committee offers no global rule for training generative AI. Instead, it identifies the decisions governments still face over permission, opt-outs, payment, transparency and cross-border remedies.
Read story
Anthropic’s public filing exposes a deliberately unusual control system: investors can buy economic exposure, while the seven founders retain majority voting power and an independent trust selects most directors. The offer size and price remain blank.
Read story
Moonshot AI removed cyber-related material from two stages of Kimi K3.1 Pro’s training before a planned open-weight release. Its own exploit test shows a real reduction, but not the elimination of offensive capability—and independent replication must wait for the weights.
Read story