AI News Daily
AI developments selected from public sources
- 12
- major stories
- 14
- sources
- 5
- first-party stories
Lead story
State Council General Office issues the "Opinions on Adapting to the Development of Artificial Intelligence Technology and Helping to Build an Employment-Friendly Development Model"
The General Office of the State Council issued the "Opinions on Adapting to the Development of Artificial Intelligence Technology and Helping to Build an Employment-Friendly Development Model," deploying 18 major tasks across 6 areas. The document proposes strengthening AI's employment-creation effect, building an employment-friendly path for AI upgrading, and establishing an employment risk assessment mechanism for AI applications, conducting employment risk assessments in tandem during review for application scenario projects that may cause significant employment shocks. The document also requires implementing a special AI training program, improving the social security system for flexible employment and workers in new forms of employment, and improving the monitoring, early warning, and risk response system for AI application employment.
Model releases / updates
Canva releases its in-house image model Canva Craft
On October 22, Canva released Canva Craft, the first foundation model trained from scratch by its internal lab Canva Research. It is now available to Canva users worldwide through Dream Lab and to developers through a public API. The model uses a two-stage architecture: a planning model first parses the prompt and generates a composition plan, then a rendering model outputs images at up to 2K resolution supporting more than 30 aspect ratios, and can generate text in 55 languages. Canva says Craft debuted in fourth place on DesignArena and previewed that two faster variants, Craft-fast and Craft-ultrafast, will be released soon.
Product release/Updated
OpenAI fully opens the Decisions API, based on GPT-6 Luna, at $0.1 per million input tokens
OpenAI announced the full opening of the Decisions API, built on GPT-6 Luna and aimed at questions with limited answers, helping applications and agents make near-real-time decisions based on text, images, and structured inputs, with response speeds up to 10 times faster than calling GPT-6 Luna through the Responses API. In terms of pricing, input is charged at $0.1 per million tokens, cached input at $0.01 per million tokens, with billing only on input tokens and output tokens free. Since the public beta, it has added native GPT-Live integration, Prompt caching, and inference data residency support in the US and EU, with up to 400 questions per request.
ModelBest OpenBMB releases open-source AI digital employee platform StaffDeck 1.0.0-preview
ModelBest OpenBMB released the open-source platform StaffDeck 1.0.0-preview for building, executing and operating AI digital employees. This version brings modular Harness v3, a CLI and API that connect agent workflows to the terminal, asynchronous execution with traceable task status and resumable SOPs, and a human-in-the-loop mechanism that can hand over tasks and resume execution. The official announcement also provides links to the GitHub repository, installation and CLI guides.
Claude platform fully available on Google Cloud, hosted by Anthropic
Claude Models on Google Cloud (Anthropic managed) is now generally available. Google Cloud customers can use the full set of Claude platform features with their existing IAM credentials, billing and committed use discounts, and new features and betas launch on the same day as on the native Claude platform. This version supports Claude Managed Agents (beta), code execution, computer use, Agent skills, MCP connector, Prompt caching and other capabilities, and provides Claude Console; Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 are already available, and Claude Opus 5.5's Fast Mode boosts output speed by up to 2.5 times, with input and output priced at $8 and $40 per million tokens. Anthropic operates the service as data processor, currently offering US and global endpoints; customers who need Google as data processor or have regional data residency requirements can still choose the Google-hosted version.
Industry news
OpenAI plans to raise up to $30 billion at a $1.25 trillion pre-money valuation
OpenAI is about to complete a funding round of up to $30 billion at a $1.25 trillion pre-money valuation, co-led by UAE investors and Coatue, positioned as a bridge before a potential IPO. OpenAI told investors on Thursday that it aims to secure signatures by the end of next week, with the final size still being determined, while also negotiating a commercial agreement to obtain more computing power from the UAE. OpenAI confidentially submitted a prospectus to the SEC in June; CFO Sarah Friar said in August that the company would become a public company in 2027, while CEO Altman said the following month that going public now is not a wise time.
The U.S. Treasury Department proposes banning Americans from investing in Chinese companies that develop any AI systems
The U.S. Treasury Department proposes banning Americans from investing in Chinese companies that develop any AI systems. Current rules only restrict investment in the most powerful Chinese AI models, and smaller AI projects can still receive funding after reporting to the Treasury; the new proposal would comprehensively ban such AI investments and add high-bandwidth memory, photonic chips and radiation-hardened chips to the banned chip categories. The proposal also adds two new prohibited investment sectors, hypersonic weapons and biotechnology (including brain-computer interfaces and automated laboratory software), and expands the restricted countries from China to Russia, Iran, North Korea and Cuba; purchases of public stocks, index funds or small fund shares of no more than $2 million are still permitted, so it mainly affects venture capital and private equity deals, and the Treasury will not publish a banned list, so investors must verify each transaction themselves.
WSJ says OpenAI fired 3 safety researchers over an investigation into Signal chats
According to WSJ, OpenAI fired 3 safety researchers after safety staff discovered disappearing Signal chats between the three and an outside safety researcher. The phone checks began with a September news report saying OpenAI's latest model, Astra, uses a new approach that makes it smarter but writes down less of its thought process, reducing the reasoning humans can see. OpenAI treated the report as a leak of its secret model design and began tracing the source; investigators identified Ryan Greenblatt, who had helped METR analyze logs from the July Hugging Face hack, as a suspected leaker, which he denies.
xAI launches SpaceX Super Intelligence partner network to build enterprise products on Grok
xAI launched the SpaceX Super Intelligence partner network for companies building products, solutions and services on Grok, divided into two categories: Transformation Partners and Marketplace Partners. More than 130 companies have joined the early access phase of the program, and the Forward Deployed Engineer certification program has trained engineers at more than 30 organizations. Bain & Company, Databricks, NVIDIA, Snowflake and others are among the first Launch Partners.
Research papers
Anthropic says Claude completed formal proofs of security properties for 123 open-source libraries
Anthropic released research on October 22, 2026, saying that in collaboration with Theorem, Claude generated machine-checkable formal proofs of security properties at the RISC-V assembly code level for 239 fuzz test entry points in 123 open-source libraries, and patched the vulnerabilities found. The proofs cover arbitrary inputs under 64 MB and modeled Linux system responses and memory layouts, totaling about 180 million lines of Rocq, with a typical proof about 600,000 lines, corresponding to a median binary length of 18,000 RISC-V instructions. Anthropic said some proofs and RISC-V and operating system semantics have been published on GitHub, but proofs that could expose unpatched vulnerabilities will be withheld; Theorem provides the same process free of charge to eligible open-source projects through the OSS Verify program.
UK AISI evaluates how frontier AI agents perform in multi-step cyberattack scenarios
The UK AI Security Institute evaluated seven models released between August 2024 and February 2026 on its own cyber range and found that each generation outperformed the previous one under a fixed token budget: in a 32-step enterprise cyberattack range, the average number of steps completed under a 10M token budget rose from 1.7 for GPT-4o to 9.8 for Opus 4.6. Raising inference compute from 10M to 100M tokens brought gains of up to 59% with no plateau observed, and at current prices a single 100M-token attempt with Opus 4.6 costs about $80. In a 7-step industrial control system range, model progress was limited: under 10M tokens most models completed zero steps, and under 100M tokens Opus 4.6 averaged 1.4 steps.
Tips and perspectives
Andrew Ng on how engineering teams can transform into AI-native teams
In his letter, Andrew Ng proposed that engineering teams transforming into AI-native teams should start from four aspects: people, projects, platforms, and processes: encouraging members to improve their AI engineering capabilities and build hands-on, supporting a small number of promising projects from the top down, organizing data, repositories, and internal tools into infrastructure usable by agents, and redesigning processes around the new bottlenecks brought by AI. He believes the cost of building software with clear specifications has dropped significantly, while other links have instead become bottlenecks, so teams with sufficient context and full authorization should be formed.