OpenAI ships Cyber; Meta opens models; Regulators move

Anthropic / Claude ecosystem

No significant new developments.

Frontier model providers

OpenAI Unveils GPT-5.6-Cyber with fewer guardrails for trusted security researchers

OpenAI has launched GPT-5.6-Cyber, a new version of its large language model with reduced guardrails, specifically designed for vetted security researchers. This expansion of the Daybreak program aims to enable defenders to test exploits more effectively while minimizing the risk of misuse by malicious actors.

Mark Zuckerberg Says Meta Will Release Weights of Muse Spark 1.2, Announces 30B Muse Glimmer Model For Local Use

Meta has announced the release of Muse Glimmer, a 30-billion-parameter open-weights agentic model optimized for local deployment on consumer hardware. Additionally, Meta confirmed it will release the weights of its frontier-capable Muse Spark 1.2 model, reaffirming its commitment to open-source AI.

AI developer tooling & infrastructure

Nvidia nemotron labs voice: NVIDIA Nemotron Labs Voice

NVIDIA has launched Nemotron Labs Voice Chat 11B, a full-duplex speech-to-speech model engineered for ultra-low latency. The model achieves approximately 450 ms turn-taking latency and supports live tool calling during speech, designed to minimize pauses and improve responsiveness in voice agent interactions.

Cloud & platform providers

Cloudflare menyatukan Workers AI dan AI Gateway: fitur…

Cloudflare has unified its Workers AI and AI Gateway into a single control plane, offering shared bindings, REST endpoints, unified billing, and integrated observability. Future plans include model-based routing and prompt classification.

AI policy, regulation & governance

Morena Unveils Constitutional Reform to Regulate Artificial Intelligence

Mexico's Morena parliamentary group has proposed a federal constitutional amendment to grant Congress explicit authority over AI legislation. This initiative aims to address existing fragmentation from seven state-level AI laws and close gaps in intellectual property protections for AI-generated works.

South Australian Premier Peter Malinauskas announces a royal commission into artificial intelligence risks and regulation

South Australian Premier Peter Malinauskas has announced a royal commission to investigate the risks and benefits of artificial intelligence and propose regulatory frameworks. This follows his recent trip to the US, where he signed a memorandum with OpenAI.

Industry & market moves

S.Korea's Naver invests $7 mn in Anthropic to shift AI strategy, expand global alliance

South Korean platform giant Naver Corp. has made a strategic investment of $7 million in US-based Anthropic. This marks Naver's first capital commitment to a top-tier global frontier AI developer, signaling a shift in its AI strategy and an expansion of its international alliances.

Israeli AI startup Decart nearing $6 billion sale, Musk’s SpaceX circling

Israeli AI startup Decart, a three-year-old video-generation AI company, is reportedly nearing a $6–7 billion sale, with Elon Musk's SpaceX among the potential acquirers. Amazon and Nebius are also cited as interested parties.

NVDA Teams Up With Wall Street Giants On Massive $500B AI Infrastructure Funding Push: Report

Nvidia is partnering with major Wall Street firms including Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR, in a massive $500 billion AI infrastructure funding initiative. This collaboration aims to cover investments in AI chips, power generation, and data centers.

AI product & feature launches

No significant new developments.

Research with immediate practical relevance

Study finds LLM-native IDE security risks in system controls

A new study by York University and the University of Calgary has identified 32 categories of security and privacy issues in LLM-assisted development environments. Unauthorized file operations represent the largest category, accounting for 43.1% of reported vulnerabilities.

Harvard and MIT Researchers Launch MatrAIx, a Simulation of 8.3 Billion AI Personas

Researchers from Harvard and MIT have unveiled MatrAIx, a large-scale simulation environment capable of hosting 8.3 billion AI personas. This platform is designed to rigorously test AI system behavior and expose ethical fault lines on an unprecedented scale.