Tech news digest for Wednesday, July 8, 2026

OpenAI's GPT-5.6 Models Set for Public Release

OpenAI is set to publicly release its GPT-5.6 models, including Sol, Terra, and Luna, on July 9, 2026, after lifting government-imposed restrictions. The U.S. Department of Commerce has approved the rollout, allowing OpenAI to expand access globally following initial testing and a collaborative review with the government.

• GPT-5.6 Sol Ultra achieved a 91.9% score in the TerminalBench 2.1 coding benchmark, outperforming Anthropic's Claude Mythos 5, which scored 88%.

Apple Partners with Broadcom for U.S.-Made Wireless Chips

Apple has entered into a significant partnership with Broadcom, signing a $30 billion multiyear deal to produce wireless chips in the United States. This collaboration is expected to yield over 15 billion custom wireless connectivity chips and aims to create hundreds of American jobs, although the total number may be relatively small.

• Broadcom will manufacture the chips in the U.S. as part of the deal.
• The partnership is part of Apple's strategy to enhance domestic production capabilities.

SambaNova Valued at $11 Billion After Funding

SambaNova has achieved an impressive valuation of $11 billion following a successful funding round that raised $1 billion. This investment was backed by notable firms including Seligman Ventures, T. Rowe Price, and Capital Group, and will support the deployment of SambaNova's systems for enterprise AI workloads, particularly utilizing their SN50 chip for on-premise AI applications.

• JPMorgan Chase plans to implement SambaNova's systems for enterprise AI workloads.
• SambaNova's SN50 chip is designed for on-premise AI deployments.

China Issues Warning Over Anthropic's AI Tool Vulnerability

The Chinese Ministry of Industry and Information Technology has flagged a significant back-door vulnerability in Claude Code, an AI tool developed by Anthropic. As a precaution, users are advised to uninstall or upgrade from versions 2.1.91 to 2.1.196, and Alibaba has mandated that employees cease using Anthropic tools starting July 10.

• Claude Code has been identified with a serious security risk, prompting immediate action from the Chinese government.

Apple's $30 billion chipmaking deal with Broadcom

Apple has announced a significant investment of $30 billion in partnership with Broadcom to bolster U.S. chip manufacturing. This deal includes a $1.5 billion expansion of Broadcom's facility located in Fort Collins, Colorado, and will ensure the supply of custom ASIC silicon products through 2031.

• Broadcom's facility expansion in Colorado is part of a broader strategy to enhance domestic chip production.

SambaNova Secures $1 Billion Funding Round

AI chip startup SambaNova has successfully raised $1 billion in its Series F funding round, achieving a valuation of $11 billion. The funding was led by General Atlantic and included contributions from Seligman Ventures and T. Rowe Price, aimed at enhancing its supply chain and expanding its AI chip offerings.

• JPMorgan Chase has selected SambaNova as an inference-infrastructure partner for secure AI solutions.
• The company previously raised $350 million in February to support the development of its SN50 AI chip.

Prime Intellect Raises $130M to Enhance Enterprise AI Development

Prime Intellect has successfully secured $130 million in Series A funding to empower enterprises in developing their own AI agents. The startup's platform is already being utilized by notable clients such as Ramp, Zapier, and Flapping Airplanes, who benefit from its tools to enhance AI capabilities while maintaining data control.

• The funding aims to democratize AI training, allowing companies to avoid reliance on closed AI labs.
• Ramp reported superior performance using Prime Intellect's AI development tools.

Google revises Android AI model evaluations

Google is updating its Android Bench rankings with the introduction of the Harbor framework for evaluating AI models. The newly top-ranked model, Claude Fable 5, achieved a score of 84.5, surpassing GPT-5.5 by 4 points. Developers are encouraged to contribute benchmark evaluations and submit tasks to enhance the ranking system.

• Claude Fable 5 is now the highest-rated AI model for Android development.
• The Harbor framework aims to improve the evaluation process for AI models.

Samsung Unveils Galaxy Z Fold 8 and Other Devices Ahead of Launch

Samsung is set to reveal its latest devices, including the Galaxy Z Fold 8, on July 22nd. Leaked renders confirm a new design featuring a 7.6-inch QHD+ inner display and a 4,800mAh battery with 25W charging speeds. The Fold 8 will be available in four colors, including a Samsung.com exclusive 'Pistachio'.

• The Galaxy Z Fold 8 will offer storage options of 256GB, 512GB, and 1TB, but will not include a 16GB RAM variant.
• The Galaxy Z Flip 8 will feature a larger cover display and color options including Graphite, Cream, and Pink.

Anthropic's fix for Fable 5's high cost is turning it into a manager that delegates to Sonnet 5

Anthropic has restructured its Claude Fable 5 model to address its high operational costs by positioning it as a planner that delegates tasks to the smaller Sonnet 5 model. This new approach includes two strategies: the 'Advisor' pattern, where Sonnet 5 executes tasks and consults Fable 5 for guidance, achieving 92% of Fable 5's performance at 63% of the cost, and the 'Orchestrator' pattern, which allows Fable 5 to manage multiple Sonnet 5 agents, delivering 96% of its performance at just 46% of the cost. This shift is likely a response to increasing price competition from Chinese models and the newly introduced GPT-5.6 Sol, which is cheaper and more efficient.

• Growing price pressure from Chinese models and GPT-5.6 Sol is influencing these changes.

Google Chrome will remove older Manifest V2 extensions in August

Google Chrome has set August 31, 2026, as the definitive end date for all Manifest V2 extensions, which have been phased out over the past four years. While these extensions can still be used if installed before the release of Chrome 138, they will no longer receive updates or be available for reinstallation from the Chrome Web Store. Google disabled support for Manifest V2 extensions in July 2025, and this final removal marks the transition to the more secure but restrictive Manifest V3 framework, which has drawn criticism for impacting ad blockers and similar tools.

• Manifest V2 extensions have been unsupported since July 2025.

The U.S. strikes Iran, Samsung's Wall Street letdown, using AI for financial planning and more in Morning Squawk

Tensions between the U.S. and Iran escalated as President Donald Trump announced the ceasefire is 'over' during a NATO summit, leading to U.S. Central Command launching a series of strikes in Iran. This development has raised fears of a broader conflict in the Middle East, particularly concerning the Strait of Hormuz. Following the strikes, oil prices jumped nearly 6%, contributing to a decline in U.S. stock markets, with the Dow Jones Industrial Average falling below 53,000. The U.S. Treasury also revoked the authorization for Iranian oil sales, further impacting market dynamics.

• Samsung's stock dropped 8% despite reporting a quarterly profit surge of 1,800%.
• Stellantis launched the all-electric Fiat Topolino in the U.S. starting at $13,995.

Former OpenAI exec Kevin Weil is now on the board of Stoke Space

Kevin Weil, a seasoned tech executive with experience at Twitter, Meta, Planet Labs, and OpenAI, has joined the board of Stoke Space, a Seattle startup developing reusable rockets to rival SpaceX. Stoke's CEO, Andy Lapsa, highlighted Weil's critical role in fundraising, helping the company secure $1.34 billion since its inception in 2020, including a $510 million Series D round in 2025. Weil's background in digital products contrasts with Stoke's focus on aerospace, raising questions about potential synergies with OpenAI, especially given previous interest from Sam Altman in Stoke's technology.

• Weil previously served as OpenAI's chief product officer and led efforts in scientific research.
• Stoke's Nova rocket aims for complete reusability, a feat not yet achieved in the industry.

Meta tests always-on AI glasses that capture your entire day

Meta is developing AI-powered glasses featuring a capability known as 'Super Sensing,' which allows the device to continuously record the user's environment through cameras and microphones. This functionality enables users to query an AI for any audio or visual information captured throughout the day. The project has raised privacy concerns, as the glasses will not have an LED indicator light to inform bystanders when recording is taking place. Additionally, Meta is considering using the data collected to enhance its AI models, building on insights from its long-running Project Aria research program.

• Meta's Project Aria has been collecting first-person data for AI systems for several years.
• The glasses are designed to assist users with tasks by building context throughout the day.

Muse Image is technically impressive, but Meta's use of Instagram photos raises questions

Meta has introduced Muse Image, its first image generation model from Superintelligence Labs, designed to function as an AI agent. Unlike traditional models, Muse Image refines its outputs by utilizing external tools, including web search, to enhance accuracy. It currently ranks second on the Image Arena platform, trailing OpenAI's GPT Image 2. However, the model's opt-out feature, which allows users to generate images using public Instagram photos without consent, has sparked privacy concerns, particularly in Europe where regulators are expected to scrutinize its compliance with GDPR and the upcoming EU AI Act.

The robotaxi law that could ban Tesla

New Jersey lawmakers are set to vote on a bill that would require companies operating fully autonomous vehicles in the state to utilize multiple sensing technologies, including lidar and radar, effectively banning Tesla's camera-only Robotaxi system. This legislation, if enacted, would make New Jersey the first state to establish such hardware mandates, ahead of a similar proposal in New York. The bill aims to ensure safety in the state's densely populated areas and includes a three-year pilot program that mandates extensive testing before companies can deploy fully driverless services.

• Tesla has lobbied against the bill, claiming it unfairly targets their technology.