Artificial Intelligence.

The European Commission has published a draft guidance document of over 130 pages to help companies determine if their AI systems fall under the high-risk category in the AI Act — but criticism from the industry is harsh.

A technical discussion on Lobsters AI states outright what many security teams don't want to hear: allowlists are not enough when AI agents can exfiltrate data through channels you forgot to consider.

OpenAI's reasoning model has disproven a geometry conjecture from 1946 — and this time, prominent mathematicians stand behind the result.

The shadow library Anna's Archive has published a direct message to AI models, asking them to donate. HN is boiling over — and the background is far more explosive than the file itself.

The hacker group TeamPCP has compromised thousands of open-source projects – including GitHub itself – in what is described as one of the most extensive software supply chain attacks in history.

The EU has revised the AI Regulation, giving companies using high-risk AI significantly more time to comply with the regulations. For Norwegian businesses operating in the EU market, this means a new deadline in December 2027.

Following a plane crash in Kentucky, people used AI tools to reconstruct the last words of deceased pilots from audio images in a public document — forcing the US National Transportation Safety Board to shut down its entire public archive.

ThunderKittens from Stanford's Hazy Research Lab crushes established tools in GPU performance — and the community is starting to notice.

While the US negotiates access to the most powerful AI systems for critical infrastructure protection, Europe is largely shut out — with only 5 percent of the world's AI computing power and rules that won't take effect until 2027.

Google's new AI search gives users answers without clicking through – and the consequences for content providers could be dramatic. Store Norske Leksikon already lost 25 percent of its traffic overnight.

A technical deep dive on Lobsters AI goes viral: TurboQuant from Google Research compresses the KV cache in LLMs to just 3 bits — without sacrificing accuracy.

A new open-source tool for stripping AI watermarks from images has hit the top of Hacker News. The debate is heated — and the questions are serious.

Google is rebuilding its search engine from the ground up with AI-powered answers and autonomous agents — and it could cost online publishers up to two billion dollars in lost advertising revenue.

Google has launched Gemini Omni Flash – a multimodal AI model that generates and edits video from text, images, and audio through natural conversation. According to Google DeepMind, this is a step towards AGI.

The EU has extended the deadline for the strictest requirements of the AI Act by over a year. Norwegian companies and researchers developing high-risk AI will have more time – but warnings about loopholes are pouring in.

The programming language that promises to make Python as fast as C++ has reached a historic milestone. The AI underground's reaction is warm — but with reservations.

Former Nav director Hans Christian Holte has been appointed head of KI Norge — the state's new initiative to make Norway a leading AI nation.

Mira Murati, former Chief Technology Officer at OpenAI, has testified under oath that Sam Altman gave her false information about the security protocols for a new AI model. The testimony emerged during the ongoing Musk v. Altman lawsuit.

Google has quietly installed a 4 GB AI model on hundreds of millions of devices via Chrome — without telling you about it. Now, Hacker News is abuzz.

In May 2026, Norway officially joined Pax Silica – a US-led coalition aimed at reducing Western dependence on Chinese supply chains for artificial intelligence and semiconductors.

OpenAI has made GPT-5.5 Instant the standard for all ChatGPT users. The model promises drastically fewer errors and more personalized responses – but competitor Google is not without its advantages.

Someone has squeezed Llama2 inference into under 1500 bytes of x86-assembly and boots it directly from disk — without an operating system. The Lobsters AI community is buzzing.

Norway is joining the U.S.-led Pax Silica initiative, which aims to secure access to critical raw materials and semiconductors for artificial intelligence development. The Ministry of Trade, Industry and Fisheries confirms its participation.

Moonshot AI's Kimi K2.6 has just been released as an open-weights model — and the HN thread is exploding after it topped a programming benchmark against the biggest proprietary players.

The US is strengthening its military AI initiative with new classified agreements with OpenAI, Google, and Nvidia – but Anthropic is banned after a bitter dispute over security limits.

A Bergen-based company announces downsizing, pointing to artificial intelligence as the driving force behind the cuts. The trend is global but is now hitting locally in Bergen.

Developers report that Claude Code behaves strangely — or simply refuses — when commit messages contain the word 'OpenClaw'. HN is buzzing.

In an ongoing lawsuit, Elon Musk claims that OpenAI's co-founders deliberately misled him about the company's future – and that he himself was naive enough to believe them.

An explosive debate on Hacker News poses a question no AI companies truly want to answer: who owns the code your AI assistant just generated?

Negotiations on exemptions in the EU AI Regulation have become deadlocked, increasing uncertainty for Norwegian and European AI companies preparing for costly compliance.

Cognition's AI engineer Devin is now appearing as a full-fledged CLI tool, and people on Product Hunt are starting to understand what that actually means for developers who live in the terminal.

NVIDIA launches Nemotron 3 Nano Omni – a compact, open multimodal AI model with a one-million-token context window that outperforms rivals from Qwen and OpenAI in inference speed.

David Silver, known from DeepMind and AlphaGo, has raised $1.1 billion for his new company Ineffable Intelligence — which focuses on AI that learns independently without human training data.

An interactive walkthrough from Lobsters AI is accelerating discussions about Google's TurboQuant — a new quantization algorithm that reportedly solves one of the most annoying memory problems in LLM inference.

Anthropic has quietly built an integration ecosystem that is starting to look serious. People on Product Hunt have begun to notice.

Phoenix Ikner asked ChatGPT several questions before shooting multiple people at a Florida university. OpenAI is now under investigation – and the case has a direct Norwegian connection.

OpenAI has launched GPT-5.5 and is working towards a unified desktop 'super app.' The company is now competing against Anthropic and Google for the enterprise market – with Norwegian developers in the crosshairs.

A blog post about canceling a Claude subscription has exploded on Hacker News with almost 600 comments – and the frustration is widespread and relatable.

New AI models make phishing attacks dramatically faster and more accurate. A security expert describes the scale attackers now operate at as the single greatest threat.

Google has committed to investing up to $40 billion in AI company Anthropic – the largest single commitment in the company's history. The investment combines cash and access to massive computing capacity.