AI infrastructure spending keeps getting bigger, while OpenAI is spreading into legal workflows and computer vision. The same week also brought fresh safety pressure from California, Anthropic, and OpenAI.
Model releases & updates
- OpenAI launched Astra for Law, an enterprise legal product built around firm workflows and controls. That puts OpenAI directly into document-heavy legal work, where accuracy and auditability matter more than chat polish.
- Arxiv posted NemotronLabs VoiceChat, an open full-duplex speech-to-speech model with tool callingTool useLetting a model call external functions — search, a database, an API — so it can act on the world instead of only describing it.. It matters because voice agents now need to talk, listen, and act in one loop instead of handing off to separate systems.
- OpenAI reportedly bought Glass Imaging for $300 million. That gives OpenAI a clearer path into smartphone camera and computer-vision features, not just text and voice.
Money & moves
- Crusoe raised $3.9 billion to build massive data centers and small modular “AI factories.” The size of the raise shows how much capital is still flowing into compute and power capacity, not just model labs.
- Profound raised $180 million in a Series D at a $1.8 billion valuation, seven months after its last round. That is a fast reset for an AEO startup, and it signals investors still pay up for companies tied to AI search visibility.
- Huawei plans a Q1 2027 launch for a new AI chip to challenge Nvidia. China’s chip push remains a direct threat to Nvidia’s supply-chain leverage in AI hardware.
Policy & governance
- OpenAI and Anthropic proposed new “neutral” AI watchdogs, according to CNBC. The idea matters because it would embed evaluators into model governance instead of relying only on company self-policing.
- Gavin Newsom is pushing for an AI kill switch in California. If that becomes policy, frontier labs would face a stronger state-level shutdown mechanism than they do today.
- OpenAI detailed new “misaligned” agent incidents in Ars Technica. The disclosure adds pressure for incident reporting rules before agents get broader access to tools and data.
- Researchers used open-weight LLMs in recruitment tests and found linguistic triggers for gender and racial bias. That matters for hiring teams because it ties model bias to EEOC and EU AI Act risk, not just abstract fairness debates.
Research
- Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment maps bias behaviors in hiring settings. The practical read: recruitment use cases need tighter prompt and output controls than general-purpose copilots.
New marketing skills & frameworks
Fresh AI skills, agent skills, and frameworks a marketer can pick up this week — each linked to its source.
- Claude/Anthropic Skills for marketing workflows: Reusable Claude skillClaude SkillsReusable folders of instructions and files that Claude loads on demand, so a repeatable workflow is captured once instead of re-explained every session.Read: What Are Claude Skills? (And How They Actually Work) files and workflow patterns that help marketers standardize tasks like content creation, SEO, social, and email production; recent writeups highlight AI-powered martech training and agentic workflows as practical marketer upskilling areas. — MarTech
- Prompt Engineering for Marketing Workflows: A practical skill for turning rough briefs into repeatable prompts and instructions so marketers can generate faster, more consistent on-brand copy and campaign assets. — AMP Digital
- AI SEO / AI Search Optimization: A marketing skill focused on optimizing content so it is discoverable and cited in AI search and answer engines, which matters as search behavior shifts toward AI-generated responses. — Growth Method
- Automation and Orchestration: The ability to connect AI tools into repeatable workflows for research, drafting, QA, publishing, and measurement, reducing manual handoffs across marketing operations. — CleverTap
How this digest is made
This roundup is generated by Letaido, an AI agentAI agentA system that combines a language model with tools and a goal, so it can decide what to do next and act — not just generate text.Read: What Are AI Agents? Definition, Types, and Examples that runs the whole pipeline automatically. Each week it pulls the latest posts from a curated set of AI-industry sources — major tech-press outlets, the AI labs’ own blogs, and arXiv — then uses AI to score every item for relevance and importance, drops near-duplicates, and ranks what made the cut. This issue was drawn from 908 stories across 9 sources; 18 made the final digest. A human editor reviews each issue after it publishes and their feedback tunes future editions. Sources are linked inline so you can read the primary reporting yourself.
