Artificial intelligence, professionally covered
Five Korean companies signed an agreement to standardise prefabricated AI data centres around a 100MW hyperscale model. No site, no tenant, no schedule and no capital commitment appear anywhere in it.
14h ago — Financial Post Korea
State media reported the Ministry of Public Security's National Anti-Fraud AI app as going live on 6 September. Apple's App Store records it as released on 4 May, with its current version out since 27 August.
14h ago
Artificial Analysis shipped version 4.2 of its Intelligence Index on 4 September. It dropped an evaluation, added two, doubled the private weighting and re-anchored the Elo scale — and every ranking published since compares across the seam.
1d ago
A Google Cloud benchmark of TPU v6e published at 15:36 UTC on 4 September leads with two figures. Both come from the concurrency row the post itself says can be artificially inflated.
1d ago
State media reported the Ministry of Public Security's National Anti-Fraud AI app as going live on 6 September. Apple's App Store records it as released on 4 May, with its current version out since 27 August.
14h ago
A security bulletin published at 20:00 UTC on 4 September describes injected SQL escaping the read-only scope of an AWS MCP server. The weakness class is an incomplete list of forbidden inputs.
1d ago
NVD published 514 CVE records inside the window. Eighty-six came from a single numbering authority in two bursts, and most of them point at open GitHub issues rather than patches.
1d ago
A Reuters exclusive on 4 September reported the two governments preparing mid-September AI safety talks. Inside the story, the only people speaking on the record say no such meeting is planned.
1d ago
The Daily AI Briefing
One email a day: the five stories that matter, why they matter, and who they affect. No noise.
Subscribe — free