AI News

Important AI releases, research, company updates, and industry developments selected from trusted sources and presented with clear context.

news

AI firms say UK has a year to head off datacentre protests before construction is threatened

Major US tech firms including Microsoft, Nvidia, and Cisco warn UK ministers that public opposition to AI datacentres is growing rapidly. They argue the UK has a critical 12-month window to establish a 'prosperity compact' offering local communities financial benefits before protests mirror those seen in the United States. While companies like Microsoft have committed billions to UK infrastructure, activist groups accuse big tech of trying to buy off communities and warn that public distrust regarding AI safety and environmental impact has already shifted against these projects.

View Details →

news

Meta to create a new class of shares to fund AI push

Meta is issuing a new class of non-voting shares to fund its aggressive AI expansion, including infrastructure, acquisitions, and employee compensation. With capital expenditures expected to reach $135-$145 billion this year, the move signals a massive financial commitment to compete in the AI race. Recent product launches like the Muse AI agent have seen rapid adoption but also faced privacy scrutiny.

View Details →

news

PHP Armageddon: OpenAI, Google’s Wiz find ancient holes, panicking the internet

Critical security vulnerabilities have been discovered in PHP, a programming language powering approximately 70% of all websites globally, including major platforms like WordPress. The U.S. National Security Agency (NSA) has issued an urgent warning, advising that unpatched sites be taken offline immediately to prevent server infiltration and database breaches. These flaws, existing since PHP version 5 released in 2004, were recently identified by AI-driven security tools from OpenAI and Google's Wiz. The discovery highlights a paradigm shift where artificial intelligence has lowered the barrier for hackers to exploit decades-old code, turning previously untargetable weaknesses into immediate threats. Consequently, the global internet community faces a race against time to update millions of websites to the latest supported PHP version (8.5.12) before malicious actors can leverage AI to adapt exploits across different configurations.

View Details →

news

Meta pauses Muse sharing after researchers find AI chats on the web

Meta has temporarily disabled the ability to share conversations with its Muse AI assistant after security researchers discovered that thousands of these links were publicly accessible via DNS records. The flaw exposed sensitive data including financial documents, calendars, and personal notes from various organizations.

View Details →

news

AI surveillance startup Flock to cut several hundred jobs amid backlash, sources say

Flock Safety, a major US-based AI surveillance company, plans to cut approximately 18% of its workforce (around 270 jobs) following intense public and political backlash over its license plate reader technology. Despite raising $950 million in venture capital and expanding into drone manufacturing, the firm faces regulatory hurdles, privacy lawsuits, and opposition from lawmakers ahead of the November midterm elections.

View Details →

news

‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop

OpenAI released nearly 400 AI-generated mathematical results across 700+ manuscripts, overwhelming the academic community. While some results show high caliber potential for top-tier journals or Fields Medals, experts warn of a 'slopocalypse' due to inconsistent formal verification in Lean, poor attribution, and unreadable write-ups. Mathematicians estimate it will take years to verify these claims, raising concerns about the future role of human researchers.

View Details →

news

Fixated on AI, the world has forgotten what is staring it in the face: nuclear annihilation | Simon Tisdall

The article argues that global attention on artificial intelligence risks overshadowing the immediate and escalating threat of nuclear annihilation. It highlights dangerous geopolitical trends, including Russia's saber-rattling, US policy shifts under Donald Trump regarding Iran and North Korea, and a general lack of international disarmament efforts. The author warns that complacency and silence in the face of rising military spending and weaponized space capabilities could lead to a catastrophic third world war.

View Details →

news

AI Is Getting Really Good at Messing With Cybercriminals

AI technology is being deployed by companies like Apate to combat cybercrime by acting as 'perfect victims' for scammers. These AI bots engage fraudsters in long conversations, wasting their time and resources while collecting intelligence on scam operations. Additionally, researchers are using LLM-powered honeypots to deceive AI-driven attackers, marking a new era where generative AI is used defensively to disrupt criminal enterprises.

View Details →

news

We’re putting too much faith in AI’s ability to say no

The article critiques the over-reliance on AI safety mechanisms that force models to refuse harmful requests. It argues that while refusal is a core safety feature, it is probabilistic and imperfect, often failing against determined attackers or enabling repression by authoritarian regimes. The piece explores the technical opacity of how models learn to say 'no' and the energy costs of complex defense layers.

View Details →

news

The maker of non-text AI model Jev valued at $7.5B just weeks after launch

TypeSafe AI has secured a massive $870 million funding round at a $7.5 billion valuation just weeks after launching its non-text AI model, Jev. Led by Andreessen Horowitz, the investment highlights rapid enterprise adoption, with one-third of Fortune 500 companies already utilizing the technology. Unlike traditional large language models that generate text, Jev operates on a transformer architecture to produce calibrated decisions and probabilities, offering significantly faster processing speeds and lower token usage for automation tasks.

View Details →