AI — Page 7

The plaintext matches Royal Navy records, though the first-decode claim lacks independent cryptographic verification.

Baseten's research arm plans to publish training and monitoring methods, though its proposed safety standard still lacks a technical specification.

The GPT-6 package starts with selected firms, 73 plugins and law-firm-built workflows; OpenAI says API access comes next.

CEO Peter Sirota's BUILD package includes a Places API preview, Traffic 2.0, Notion and Figma connectors, and a command-line tool.

The agent's voice mode is rolling out on desktop and mobile, five weeks after Grok Bot launched with persistent cloud computers.

The framework loads commands on demand, normalizes safety metadata and supports agent-held, host-supplied or vault-backed credentials.

BlueLedger is starting paid issuer pilots with backing from MaxWave, though customer names and investment terms remain private.

AgentCard pairs one-time-use card credentials with a stablecoin wallet, extending Alchemy's push from blockchain infrastructure into agentic commerce.

The October 1st code lets hospitals track real-time ultrasound navigation, giving Will Mauldin's Accuro 3S a route to build reimbursement evidence.

Jiacheng Miao and James Zou want journals to publish executable agents beside papers, code and data, with researchers still judging the conclusions.

HubSpot and Shopify integrations give advertisers external campaign management, attribution and product-catalog connections as OpenAI tests conversational sales agents.

Liam Fedus and Ekin Dogus Cubuk built custom routing, sandboxing and training systems for long scientific RL runs.

Bloomberg says 10 banks are providing financing for Benjamin Treynor Sloss's TPU cloud, four months after Blackstone committed $5B in equity.

Steven Liu's framework deal covers development-stage wind and solar assets; ABO Energy expects a definitive purchase agreement after due diligence.

Dan Lahav and Omer Nevo's lab found a Qwen3.5 agent could fine-tune and deploy the shared model, carrying synthetic secrets with it.

Mozilla selected Mistral Small 4 for Smart Window beta in the US, Canada and France, while preserving users' ability to switch models.

The Meta CEO says legal exposure, market incentives and outside evaluators give each frontier lab enough reason to slow on its own.

Meta employee Matt Schlicht used Muse to contact Tesla dealerships, and founder Matt Van Horn gained access to the unreleased test after saying his agent could not do the same.

The Center for AI Safety benchmark spans 10 task categories, though its public repository still says the code is coming soon.

The API release tops two independent voice benchmarks, while its 8.83-second response time leaves StepFun with a latency problem.

The MIT researcher and a distributed contributor group built an Apache-2.0 stack that updates model weights and agent harnesses from live experience.

The three-endpoint product extracts design systems, searches reference brands and scores agent output against a target identity.

GPT-6 Astra led dollars spent last week, while cut-price GPT-5.6 Luna dominated token volume across the two model families.

The screenless band is due to ship in December, while Persona's existing iPhone assistant handles email, calls, calendars and purchases.