# Uudised

See voog koondab avalikult kättesaadavaid artikleid kolmandate osapoolte allikatest. Kokkuvõtted genereeritakse automaatselt asjakohaste teemade esiletõstmiseks; need ei ole Ando Kivilaidi originaalreportaaž. Täieliku artikli lugemiseks ava allikas. Valik prioriseerib AI süsteeme, kindlustustehnoloogiat ja ehitajatele olulist inseneriuudist.

## OpenAI safety chief Johannes Heidecke to leave after reorg

OpenAIs head of safety systems, Johannes Heidecke, told colleagues he will leave the company by July 24 as safety work is folded into research under VP Mia Glaese. He is the sixth senior safety leader to exit OpenAI in two years.

[Source](https://www.engadget.com/2212941/openai-head-of-safety-leaving-company-reorganization/)

## US eases export controls, clearing UAE path for license-free AI chip sales

The US Commerce Department reclassified the UAE into Country Group A:5, effective July 10 with provisions taking effect July 14, allowing eight named US companies, Amazon, Google, Meta, Microsoft, OpenAI, Apple, Oracle, and xAI, plus approved UAE entities like G42 and Core42 to receive advanced AI chips and servers without individual export licenses.

[Source](https://9to5mac.com/2026/07/10/us-eases-restrictions-on-apples-access-to-ai-chips-and-data-center-equipment-in-the-uae/)

## Apple sues OpenAI alleging trade secret theft over AI hardware plans

Apple filed a lawsuit against OpenAI in Northern California federal court alleging systematic trade secret theft tied to OpenAI's push into consumer AI hardware, citing more than 400 former Apple employees now at OpenAI. The complaint names OpenAI hardware chief Tang Tan, a former Apple VP, and engineer Chang Liu, who allegedly retained an Apple laptop with confidential technical documents after joining OpenAI. OpenAI has denied the allegations.

[Source](https://www.cnbc.com/2026/07/10/apple-openai-lawsuit-trade-secrets.html)

## Patra Secures US Patent for AI Policy Value Extraction

Insurance outsourcing and AI firm Patra was granted US Patent No. 12,639,972 for its Policy Checking AI, which scans policy documents, extracts and maps values, and flags items for human reviewer verification. The patent, issued in May 2026, underpins Patra's governed, expert-supervised AI positioning.

[Source](https://fintech.global/2026/07/10/patra-secures-us-patent-for-ai-value-extraction/)

## Earnix launches ML-to-rating-structure distillation for transparent pricing

Insurance pricing vendor Earnix added a Model to Rating Structure Distillation capability to its Price-It platform that automatically converts complex machine learning pricing models into transparent, production-ready rating structures. The tool generates multiple candidate structures under business constraints such as monotonicity and interaction limits, letting insurers balance predictive accuracy with regulatory interpretability requirements.

[Source](https://fintech.global/2026/07/10/earnix-bridges-the-gap-between-ml-accuracy-and-regulation/)

## UK City skills compact enlists Lloyd's and 16 major firms for AI retraining push

The UK government launched a City skills compact requiring financial-services employers to run structured AI retraining programmes, with Lloyd's of London joining 16 other major firms, including Standard Chartered, Nationwide, and the London Stock Exchange, as founding signatories covering roughly 500,000 City workers. Each firm must build rolling three-year AI training plans, appoint a senior executive to oversee the programme, and report progress by November 2026.

[Source](https://www.insurancebusinessmag.com/uk/news/technology/ai-hiring-surge-drives-city-push-to-close-insurance-skills-gap-581964.aspx)

## OpenAI Launches GPT-5.6 Model Family: Sol, Terra and Luna

OpenAI began rolling out GPT-5.6 in three tiers: flagship Sol for complex reasoning and coding, balanced Terra at roughly GPT-5.5-level performance for half the cost, and fast, low-cost Luna. The family adds programmatic tool calling in the Responses API and ships across ChatGPT, Codex and the OpenAI API over 24 hours.

[Source](https://www.marktechpost.com/2026/07/09/openai-releases-gpt-5-6-a-three-tier-model-family-with-programmatic-tool-calling/)

## Meta Launches Muse Spark 1.1, Its First Paid Frontier Model API

Meta Superintelligence Labs released Muse Spark 1.1, a multimodal reasoning model with a 1-million-token context window built for agentic and coding tasks. It is Meta's first model available on a paid developer API, priced at $1.25/$4.25 per million tokens, marking Meta's entry into charging for frontier model access like OpenAI and Anthropic.

[Source](https://techcrunch.com/2026/07/09/meta-enters-the-crowded-ai-coding-battle-with-muse-spark-1-1/)

## NYT-Led Publishers Ask Court to Sanction OpenAI Over Withheld Chat Log Evidence

The New York Times, New York Daily News and other publishers asked a federal judge to sanction OpenAI in their copyright lawsuit, alleging OpenAI misrepresented its ability to search training data and chat logs for copyrighted journalism despite having built a database of tens of millions of de-identified ChatGPT conversations. The publishers want the court to bar OpenAI from relying on its evidence sample and cover related legal fees.

[Source](https://techcrunch.com/2026/07/09/new-york-times-says-openai-hid-evidence-in-chatgpt-copyright-trial/)

## Fidji Simo Steps Down From OpenAI's No. 2 Role Over Chronic Illness

Fidji Simo, OpenAI's CEO of Applications and de facto second-in-command since May 2025, announced she is stepping back from full-time duties to focus on recovery from a severe POTS flare-up, transitioning to a part-time advisor role. Her responsibilities are being split among President Greg Brockman, CFO Sarah Friar and Chief Strategy Officer Jason Kwon.

[Source](https://techcrunch.com/2026/07/09/fidji-simo-steps-down-from-openais-no-2-role/)

## BMO Insurance launches AI-enhanced SmartDecision underwriting tool

BMO Insurance launched SmartDecision, an AI-enhanced underwriting tool that delivers real-time coverage decisions up to $5 million across Term, Universal and Whole Life policies, eliminating traditional medical requirements for many applicants.

[Source](https://www.newswire.ca/news-releases/bmo-insurance-launches-ai-enhanced-smartdecision-tool-delivering-a-faster-and-more-seamless-path-to-coverage-for-clients-and-advisors-805805551.html)

## Roadzen to acquire European car-rental insurance MGA Riverside

Nasdaq-listed Roadzen signed a definitive agreement to acquire European car-rental insurance MGA Riverside for about $15 million, with plans to apply its AI computer-vision claims assessment and real-time dynamic pricing technology to roughly 800,000 policies written annually.

[Source](https://www.globenewswire.com/news-release/2026/07/09/3324715/0/en/roadzen-signs-definitive-agreement-to-acquire-a-leading-european-mga-focused-on-short-term-car-rental-insurance.html)

## Metas new in-house AI chips to begin production in September

Meta is on track to start manufacturing the latest version of its custom AI chip in September, aimed at cutting GPU costs amid an ongoing component shortage, according to a Reuters report.

[Source](https://techcrunch.com/2026/07/09/metas-new-ai-chips-will-begin-production-in-september/)

## Humain and Cohere partner on Middle East AI infrastructure

Saudi Arabias Humain and Canadas Cohere announced a partnership, unveiled during Canadian PM Mark Carneys Saudi visit, with Humain committing at least 50MW of AI compute to power Cohere models and joint work on sovereign AI. Its Cohere first large compute deployment outside North America.

[Source](https://fastcompanyme.com/news/humain-and-cohere-partner-to-build-one-of-the-middle-easts-largest-ai-infrastructure-deployments/)

## Cerebras and OpenAI Detail $20B-Plus Compute Deal With New European Data Centers

At the RAISE Summit in Paris, Cerebras and OpenAI executives outlined an expanding infrastructure partnership worth over $20 billion, including a build-out of 200MW of AI compute capacity across Europe (Lyon, Norway, Finland) by end of 2027. The deal is framed as a direct challenge to Nvidia's dominance in AI compute.

[Source](https://www.euronews.com/business/2026/07/09/cerebras-targets-europe-with-multibillion-dollar-ai-expansion-challenging-nvidia)

## OpenAI Launches ChatGPT Work, an Agent for End-to-End Office Tasks

OpenAI introduced ChatGPT Work, a GPT-5.6-powered agent that researches, works across connected apps and files, and produces finished documents, spreadsheets, presentations and reports rather than just answering questions. It began rolling out July 9 to Pro, Enterprise and Edu users, with Plus and Business access following within days.

[Source](https://www.bnnbloomberg.ca/business/artificial-intelligence/2026/07/09/openai-launches-chatgpt-work/)

## ZestyAI Expands Partnership With DUAL North America for California Wildfire Cover

DUAL North America expanded its partnership with ZestyAI, deploying the Z-FIRE AI wildfire risk model for a new California homeowners program. Z-FIRE scores property-level features like defensible space, vegetation and construction materials, and is the first AI-driven wildfire model cleared in a California carrier rate filing.

[Source](https://insurance-edge.net/2026/07/09/zestyai-expands-partnership-with-dual-north-america/)

## Applied and Travelers Roll Out Submissionless Insurance Quoting

Applied Systems launched a submissionless commercial insurance quoting capability, powered by the Cytora agentic AI platform it acquired in 2025, with Travelers as its first carrier partner. The tool analyzes an agency's renewal book, converts risk data into structured form and routes it directly into carrier quoting systems inside Applied Epic.

[Source](https://fintech.global/2026/07/09/applied-and-travelers-explore-submissionless-insurance-quoting/)

## xAI Launches Grok 4.5, Its First Model Built with Cursor

xAI publicly launched Grok 4.5, a roughly 1.5-trillion-parameter model built on its V9 foundation and trained in partnership with Cursor, targeting software engineering, legal and financial-analysis use cases rather than general chatbot use. Priced at $2 per million input tokens and $6 per million output tokens, xAI claims performance roughly comparable to Anthropic's Opus 4.8 on some benchmarks at lower cost and higher speed.

[Source](https://thetechportal.com/2026/07/09/spacexai-launches-grok-4-5-its-first-ai-model-built-with-cursor-featuring-advanced-legal-and-finance-capabilities/)

## hyperexponential launches hyperoperator, a unified agent for underwriting work

hyperexponential introduced hyperoperator, an end-to-end agentic underwriting workbench for commercial P&C insurance that moves broker submissions from intake to a decision-ready file within a carriers pricing and authority controls. The platform already supports underwriting for over $75bn in premiums.

[Source](https://www.manilatimes.net/2026/07/08/tmt-newswire/globenewswire/hyperexponential-introduces-hyperoperator-the-unified-agent-for-underwriting-work/2380806)