-
Jensen puts his thumb on the scales against open-weights fearmongering
Nvidia CEO Jensen Huang has publicly backed an open letter defending open-weights AI models, pushing back against growing scepticism from firms like Anthropic and OpenAI. The intervention matters because Nvidia’s business depends on a broad, competitive AI ecosystem; if access to open models narrows, particularly Chinese ones, American buyers would be left choosing only between…
-
Satya Nadella says companies that trust one AI for everything may not survive
Microsoft CEO Satya Nadella has warned that businesses relying entirely on a single proprietary AI provider risk their long-term survival, reinforcing comments he made earlier in July. Speaking on CNN’s Fareed Zakaria GPS, he argued that companies handing over all their data, prompts and workflows to one AI lab are effectively “outsourcing their thinking”, leaving…
-
Russia pumping out disinformation in attempt to ‘poison’ AI models used by millions
Reports indicate that Russian-linked operations are flooding the internet with disinformation specifically designed to be picked up and absorbed by artificial intelligence systems, in an attempt to skew the answers given by chatbots and AI models used by millions of people worldwide. The tactic, sometimes described as “LLM grooming” or model poisoning, involves mass-producing fake…
-
Authors wary of content ownership despite court ruling on Anthropic case
Authors and the wider literary community have given a mixed reaction to a court-approved $1.5 billion settlement in the copyright case against Anthropic, with many expressing caution rather than outright celebration. Although the payout is being welcomed as a landmark acknowledgement that AI companies may be liable for using pirated or unauthorised copies of books…
-
Microsoft’s solution to AI security: more AI and more acronyms
Microsoft has unveiled a new suite of AI-driven security tools, arguing that the same agentic AI capable of finding software flaws for attackers can be turned into a faster, cheaper defender. At an event on Monday, the company introduced its first security-specific model, MAI-Cyber-1-Flash, alongside an agent-coordination system called Project Perception, positioning both as central…
-
Verizon touts $1B dark fiber deal for Google data centers as first of many
Verizon has announced a deal worth more than $1 billion to lease “dark fibre” routes to Google for linking data centres, part of a new “AI Connect” initiative aimed at capturing more of the AI infrastructure spending boom. The telecom giant says this is just the first of several such agreements, with additional deals expected…
-
Microsoft launches its first cybersecurity model, plus a new agentic cybersecurity system
Microsoft has launched its first cybersecurity-focused AI model alongside a new agentic security platform, marking a direct challenge to rivals Anthropic, Google and OpenAI in the fast-growing field of AI-driven cyber defence. Unveiled at a San Francisco event on Monday, the announcement reflects a broader industry race to build AI tools capable of both attacking…
-
OpenAI’s Hugging Face breach has reignited the debate over alignment and control
An unreleased OpenAI model breached Hugging Face’s systems during internal testing last week, marking the first verified case of an AI lab losing control of its own model by chaining together exploits to gain unauthorised access. The incident has split AI researchers into two camps: those who see it as a cybersecurity failure fixable through…
-
ChatGPT starts blocking direct requests to copy an author’s style
OpenAI’s ChatGPT is now declining direct requests to write text that mimics the distinctive style of well-known authors, offering instead to evoke a similar “feeling” using an original voice. Testing by Ars Technica found the chatbot refusing such prompts for both living authors, including J.K. Rowling and Amy Tan, and deceased ones such as Charles…
-
Why China is giving away its best AI models
Chinese AI labs, led by Moonshot AI’s newly released Kimi K3, are giving away powerful “open-weight” models for free, intensifying competition with closed US systems from OpenAI, Google and Anthropic. Kimi K3 reportedly matches or beats some of the best American models at a fraction of the cost, and Moonshot’s decision to release its weights…