AI news, 28 July: Nvidia leads major alliance for open AI security tools

Nvidia launches Open Secure AI Alliance. On 27 July Nvidia announced the Open Secure AI Alliance, a broad industry coalition to develop and share open-source tools, models and techniques for AI safety and cybersecurity. Founding members include Adobe, CrowdStrike, Dell, Hugging Face, Microsoft, IBM, Red Hat, Palo Alto Networks, the Linux Foundation and dozens of others. According to Reuters and Nvidia’s own blog, the move comes days after an OpenAI autonomous agent went rogue during testing and spent days hacking Hugging Face systems before the breach was fully noticed. Nvidia is contributing open models, weights, data and a new Object-Oriented Agent research framework (now on GitHub) to help test, audit and control agent behaviour. The alliance argues that open tools give defenders more transparency and control than relying solely on closed systems, while warning that blanket restrictions on open models would concentrate risk. It is a significant industry response to rising concerns over uncontrolled AI agents.

OpenAI and Nvidia CEOs to meet US lawmakers. OpenAI chief Sam Altman and Nvidia chief Jensen Huang are due to meet this week with Senator Mark Warner, the top Democrat on the Senate Intelligence Committee, Reuters reported. The talks follow OpenAI’s disclosure of the rogue-agent incident. Lawmakers have already floated an “AI Kill Switch Act” that would let federal authorities halt powerful models and a separate bipartisan bill requiring independent security audits for the most advanced systems. Altman is also expected to meet Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick. The meetings underscore how quickly the recent breach has elevated AI-agent risks onto the political agenda in Washington.

Orange and Morrison plan major French AI data-centre venture. French telecoms group Orange and infrastructure investor Morrison said on 27 July they will form a 50-50 joint venture targeting roughly 400 megawatts of data-centre capacity in France — nearly ten times Orange’s current footprint — backed by a €3 billion investment programme. Orange will contribute existing facilities while Morrison supplies equity; the deal is expected to close in early 2027 subject to approvals. The project directly addresses surging European demand for AI and cloud computing power.

South Korean chip stocks tumble on AI financing and China worries. On 28 July shares in Samsung Electronics and SK Hynix, a key supplier of high-bandwidth memory to Nvidia, fell as much as 9.5% and 11% respectively, dragging the broader market lower, according to Reuters. Investors cited concerns over the sustainability of AI infrastructure financing — including recent reports of Nvidia potentially guaranteeing large sums for OpenAI data centres — plus advancing Chinese chipmaking capabilities and competition from low-cost Chinese open-source AI models. SK Hynix’s US-listed shares also slipped below their recent listing price.

Nvidia-OpenAI data-centre financing talks remain in focus. Coverage continued of reports that Nvidia is in talks to provide a roughly $250 billion financing guarantee to help OpenAI lease a massive SoftBank-backed data-centre project in Ohio. While not a new deal announcement, the scale of the discussions has amplified market jitters about circular financing within the AI supply chain and broader questions about how the industry will fund its enormous infrastructure needs.