A ShinyHunters affiliate breaches a SaaS provider, steals data from roughly 200 customer organizations, and releases more than 2,100 Azure AD token sets spanning over 40 tenants in about 34 hours.
Anthropic says a China-aligned account lacking Arabic skills uses Claude over several days to draft regional-dialect messages recruiting Uyghurs in Syria and translate targets’ replies live, while separate suspected Russian and Yemen-linked users employ automated workflows for phishing and missile-related coding.
Z.AI raises fresh capital by offering 21.97 million shares at HK$714—10% below the prior close—and selling 20.14 billion yuan of zero-coupon bonds due in September 2027, some carrying yields as low as minus 0.5%.
A local code-analysis system now pairs token search and graph expansion with optional LanceDB vectors, lifting reported recall from 58.2% to 71.4% while its SHA-256 ingest queue skips unchanged files and retries failures up to three times.
OmniVoice now generates 24 kHz speech across 600-plus languages from 3–10-second voice samples, though cross-lingual clones retain the accent of the reference speaker’s language.
OpenAI pauses new $200-a-month Pro subscriptions, citing the tier’s heavy infrastructure demand, while keeping API, Go and Plus plans available ahead of Astra’s September 3 rollout to Pro, Plus, Enterprise and Business users.
Speak launches GPT-Live-1 at $0.05 per minute, claiming its front-end voice layer cuts learner interruptions during thinking pauses by nearly 80% and boosts Full Duplex Bench scores by 30 points over GPT-Realtime-2.1.
A6000 hits all 100 held-out stationary steering targets after 600 optimizer updates in 186 seconds, but a 24-second CARLA test shows that disconnecting its simulated foot grip wipes out more than 99% of steering response while visual driving and pedal control remain untested.
Google releases the vendor-neutral google-cloud-developer plugin, giving coding agents gcloud CLI guardrails, authentication and project-management guidance, and Developer Knowledge MCP access to current official documentation through Antigravity CLI, Claude Code, and Codex CLI.
WearableQA launches 4,084 ten-option health questions from 200 users’ wearable records spanning up to roughly 500 days each, testing models on 28-day windows that can include 17 blood biomarkers while showing that serialization choices shift accuracy by several points.