AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
The benchmark confronts agents with design tasks and evaluates them on existing experimental measurements. He also studies the effect of specialized tools.
A study of 18 model pairs places recognition more in the attention and refusal decision in the MLP. Sentinel combines both signals.
Electrical networks, water, transports: Anthropic joins CrowdStrike, Palo Alto Networks, Dragos or Deloitte to correct faults with Claude, and launches a free open source scanner.
Peruvian and Ecuadorian schools booby-trapped, fake audios, a façade research center: OpenAI details a campaign linked to the Politology network, and an Iranian operation of fake journalists.
Directories left open by the attacker reveal a Chinese agentic pentest tool plugged into DeepSeek, GLM, and Grok. CrowdStrike leans toward a Chinese-speaking actor motivated by money.
Following OpenAI’s flurry of mathematical findings, Ethereum researchers fear that AI could undermine elliptic curves even before the advent of quantum computing, and Scott Aaronson claims that laboratories are already testing their models in this area.
In a letter addressed to the board of directors, Tomek Korbak, Mikita Balesni and Jasmine Wang have raised concerns about the loss of transparency in the ‘thought processes’ of models such as GPT-6 Astra.
The group says it took action against 33.2 million pieces of child sexual exploitation material in the first half of the year and has introduced an AI agent tasked with bypassing its own safeguards.
Previously available only to journalists and researchers, the tool recognises the SynthID watermark used by Google, as well as by OpenAI, Nvidia and Kakao.
Defence, red teaming or specialised access: verified security teams are granted less restricted access. Glasswing claims to have identified over 129,000 confirmed vulnerabilities.
Anthropic’s model demonstrated that it was possible to reconstruct the file server’s session keys from the output of Math.random(). Attacks followed within 24 hours.
Since 1 October, the OSS Vulnerability Rewards Programme has been put on hold: the flood of automated reports, the vast majority of which were invalid, was overwhelming the engineers and maintainers.
Users were constantly asking on Amap for directions to the various entrances to parks, zoos and hospitals. The activity appears to be aimed at circumventing the API’s restrictions rather than causing harm.
The Dutch offensive security firm has raised its total funding to $65 million. Its AI agents map the attack surface and verify which vulnerabilities are actually exploitable.
After intrusions into Shinhan, KB Kookmin, Hana and Woori, Lee Jae Myung evokes "signs" of AI use and calls for cybersecurity adapted to this new era.
Researchers have put AI models through their paces against previously unknown vulnerabilities: the defence prevails over the attack in two out of five environments, and a validated patch is not enough to block everything.
The Internet Watch Foundation assessed 6,310 AI-generated images that fell within the legal definition of child sexual abuse material in the first half of 2026, compared with just over 4,500 for the whole of the previous year.
Fewer than 40 per cent of users of programming agents isolate them in a sandbox. A researcher has proposed SideKernel, a local, open-source micro-VM designed first and foremost to be easy to use.
Researchers have criticised the proposed conditional labelling mechanisms for identifying phishing content generated by open LLMs: their ‘Untag’ framework renders them completely ineffective.
Training data is encrypted on the phone and can only be read by audited code running in secure environments. Gboard is already using it.