Frontier model releases, benchmarks, and price changes — in one live feed.

Crunchbase compared recent startup funding in Argentina and Spain ahead of the World Cup final, noting that Argentina’s ecosystem remains “small, scrappy and sometimes very successful” while Spain still trails in venture dollars despite a busy pipeline. Argentina has seen 2026 funding already surpass last year with big fintech rounds like Pomelo’s $55 million Series C and Tapi’s $27 million Series B, while Spain has raised less than $2 billion so far this year, led by PLD Space’s $206 million Series C and Factorial’s $150 million Series D at a $2.5 billion valuation.

v2.46.0, dated 2026-07-17, adds the `/organization/projects/{project_id}/service_accounts/{service_account_id}/api_keys` endpoint and an `owner_project_access` field to `APIKeyListParams`, alongside several API manual updates. It also fixes generated type compatibility and removes beta-annotation compatibility aliases, signaling a cleanup of the API surface for downstream clients.

v0.117.0, dated 2026-07-16, adds API support for dreaming and MCP Tunnels, with bug fixes and documentation updates in the same release. The notable technical change is that credential material is kept out of traceback frame locals via `SecretStr`, improving secret handling during exceptions.

A VentureBeat Pulse Research survey of 107 enterprises found that 54% have already had an AI agent security incident or near-miss, while only 32% give every agent its own scoped identity and 30% isolate their highest-risk agents. The gap matters because most agents still share human, service-account, or API-key credentials, so a single compromised agent can expand the blast radius across systems and data.

New York Governor Kathy Hochul said her team is using AI to analyze “every single rule, regulation, [and] policy” in the state to identify outdated laws, citing examples like a $25 dog-hunting fee and a permit requirement for pregnant people working after midnight. She said the review could have taken five years at the staff level, underscoring how governments are starting to use AI for large-scale policy cleanup even as New York considers restricting new AI data centers.

A VentureBeat Pulse Research survey of 107 enterprises found that AI infrastructure spending is rising faster than organizations can measure it, with only 21% running AI in production at scale, 83% reporting GPU utilization at 50% or less, and just 44% rigorously tracking compute costs. The biggest planned investment area over the next year is AI-specialized clouds at 45%, while 64% expect to switch or add an infrastructure provider within 12 months, driven mainly by integration (41%) and total cost of ownership (35%) rather than token pricing.

Across 157 enterprises, 50% said they deployed an agent or LLM feature that passed internal evaluations but still caused a customer-facing failure, only 5% fully trust automated evaluation, and 66% already allow or are engineering toward zero-human-in-the-loop deployment for low-risk agents. The gap matters because the most-cited weakness is poor alignment with real-world outcomes (29%), while the evaluation stack remains fragmented: model providers’ native evals and having no dedicated tooling are tied at 17% each, and only about a quarter run real-time quality checks on live traffic.

NVIDIA’s Nemotron 3 Embed has ranked #1 overall on the RTEB benchmark, signaling a new top-performing embedding model for retrieval. The result matters because stronger embeddings improve agentic retrieval pipelines, where models need to find and use relevant information more reliably.

Google Vids added Gemini Omni and personal avatars to make it easier to create, edit, and star in videos. The update matters because it lowers the barrier to producing polished video content by letting users generate and personalize videos with Google’s multimodal AI.

OpenAI is making ChatGPT safer for teens by adding age-appropriate protections, learning tools, parental controls, and partnerships with experts. These changes matter because they aim to let teens use AI with stronger safeguards tailored to their age and needs.