The day in tech.

Apple's foldable iPhone will be named iPhone Duo and start at $2,000 with 256GB of storage, with higher tiers reaching around $3,000. Initial shipments arrive as early as October, and the device comes in white and dark blue.

Jacob Coxon resigned from Anthropic, warning that AI firms are racing toward self-improving superintelligence they cannot control. He spent three years in pre-training research at OpenAI and Anthropic, and advocates a temporary ban on improving model capabilities.

OpenAI Group PBC launched ChatGPT Images 2.5, an update to ChatGPT Images 2.0 that debuted in April. The new tool generates images faster with more natural lighting, richer textures, sharper details, and more refined edits.

Meta launched Muse, its first consumer AI agent, in the US. The product marks Meta's entry into the consumer AI agent market, expanding its artificial intelligence offerings beyond existing platforms and tools.

OpenAI announced its agents solved a Millennium Prize Problem, but the milestone is overshadowed by accusations it used NYU mathematician Tristan Buckmaster's and Anthropic employee Levent Alpöge's AI-assisted work without credit. OpenAI denied the accusations.

Microsoft's Patch Tuesday release addressed 974 CVEs, a record, including two exploited bugs, CVE-2026-81963 and CVE-2026-85880, confirmed by CISA. Federal agencies must patch by September 22. Adobe issued 10 bulletins covering 172 CVEs, including a zero-day in Magento and a critical Adobe Commerce bug.

Qualcomm announced a contract to supply Amazon with AI chips and networking hardware. The broader partnership includes a stock warrant enabling Amazon to buy up to 25 million shares worth $4 billion.

Anthropic withheld its latest AI model from the UK testing agency. The decision affects the agency's ability to evaluate the model's safety and capabilities. The move raises questions about voluntary AI oversight arrangements.

An AI-driven borrowing boom is reshaping Switzerland's credit market, increasing demand for loans and altering lending dynamics. The shift reflects growing reliance on artificial intelligence in financial decision-making, prompting Swiss lenders to adapt their strategies to accommodate the surge in AI-assisted borrowing activity.

OpenAI confirmed GPT-6 Astra is its first broadly deployed model to reach the Critical level for cybersecurity capabilities. The model can identify and develop functional zero-day exploits in hardened real-world critical systems without human intervention, and OpenAI is disclosing two previously unknown vulnerabilities Astra discovered.