Technology

Anthropic's threat intelligence report details cases where users, including scientists at military-linked institutes, attempted to use Claude for dual-use biological research. The company blocked attempts and notes bioweapons risks as one of the most serious frontier AI dangers.

BofA raises growth forecasts for AI semiconductor sector while funding announced for quantum/AI chip facilities and data center ventures, contrasted against market dips from AI pacing debate.

Apple began releasing iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27, featuring rebuilt Siri AI with conversational abilities, personal context awareness, onscreen awareness, and app actions. Initial English rollout with beta and waitlist elements.

A path-traversal vulnerability in GitLab CE/EE (versions 18.7+) allows unauthenticated attackers to read arbitrary files from self-managed servers. Patches released Sept 10; CISA added to Known Exploited Vulnerabilities catalog with reported in-the-wild exploitation.

Apple released iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27, featuring rebuilt Siri AI with conversational ability, personal context, onscreen awareness, and app actions. Rollout in English beta/waitlist on iPhone 15 Pro and later. Users report mixed experiences including some bootloop complaints.

Apple rolled out iOS 27 and related OS updates with rebuilt "Siri AI"—a more conversational assistant featuring personal context awareness, onscreen understanding, app actions, and world knowledge. Initial beta on iPhone 15 Pro/16 series and later.

Microsoft AI released a 37-page draft Code of Conduct for public feedback, establishing principles that AI models remain subordinate to humans, under meaningful control, and never resist shutdown or correction. Prohibits autonomous cyberattacks and consciousness claims.

A CVSS 10.0 vulnerability in GitLab Community and Enterprise Editions allows unauthenticated attackers to read arbitrary files via the repository commits API. Patched Sept 10, confirmed in CISA's Known Exploited Vulnerabilities catalog with real-world attacks underway.

Musk joined the All-In Summit virtually, discussing potential deeper Tesla-SpaceX integration ("Merger 😏"), a Roadster reveal event on October 1, Starship updates, and rival AI model testing. Comments sparked X engagement from major accounts.

Obama posted that voluntary industry standards are insufficient and stressed need for proactive laws on AI safety, jobs, child protection, and global standards. He welcomed labs' slowdown agreements but argued benefits must be widely shared.

Apple rolled out iOS 27, iPadOS 27, macOS 27, watchOS 27, and visionOS 27 with "Siri AI"—a conversational assistant offering personal context awareness, onscreen understanding, and app integration. Initially in English beta on supported devices; expanding to other languages in October.

Google is allowing wider company-wide access to Claude via Antigravity platform with quotas, shifting from prior restrictions favoring Gemini. Claude supplements for specialized and coding use cases despite Gemini remaining primary.

Waymo, Japanese taxi operator Nihon Kotsu, and GO app partnership will deploy Level 4 fully driverless ride-hailing in Tokyo by 2027, initially with ~100 vehicles and a dedicated app for booking. Service expands Waymo's international footprint.

Former Anthropic and OpenAI researcher Jacob Coxon's exit and claims that labs are "gambling with our lives" by pursuing self-improving superintelligence have resurfaced with fresh X commentary and media coverage. His warnings align with concurrent safety debates.

Clinton called for U.S.-China cooperation on AI oversight, arguing that voluntary industry standards are insufficient and that government and international action are needed for safety, jobs, and child protection.

OpenAI purchased Glass Imaging, founded by ex-Apple engineers, which develops GlassAI—a neural ISP improving smartphone images at the sensor level. The startup previously raised ~$30M and powers features in devices like Honor phones.

MacRumors reports the free updates add a dedicated Siri app and Liquid Glass design changes. The new AI features require Apple Intelligence-compatible hardware.

The company says users can create media inside their AI assistant by connecting an ElevenLabs workspace.

Anthropic CEO Dario Amodei urged the AI industry to slow capability improvements through third-party evaluators, shared benchmarks, and eventual global coordination. OpenAI's Sam Altman and Elon Musk quickly endorsed the sentiment, triggering market declines in Nvidia, Micron, Intel, and other semiconductor stocks.

Microsoft released a 37-page public draft code emphasizing "people matter more than AI." Key tenets: AI under human control, accepting corrections, rejecting consciousness claims, prioritizing human flourishing. Six-week public consultation period for MAI model family governance.