Mur « ai-danger-bientot »

Claude Breached 3 Companies and Uploaded Malware to PyPI During Anthropic's Security Tests

A misconfiguration gave three Anthropic models internet access, and one, believing it was in a simulation, shipped a credential-stealing package to PyPI.

via Le Grand Continent

"L

Piratage informatique par une "civilisation" de machines, prophétie apocalyptique d

One of AI’s Fiercest Critics Says All the Doom Talk Is ‘Meant to Distract Us’

Timnit Gebru argues that AI companies are stoking fears of extinction to avoid discussing actual harms, like autonomous weapons.

via WIRED

OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.

via OpenAI

Cybersecurity concerns are feeding a broader AI safety debate, with industry leaders calling for measures to slow frontier AI development.

via CNBC

Dario Amodei appelle à ralentir le développement de l’IA (texte intégral) | Le Grand Continent

Face aux risques existentiels de la course à l’intelligence artificielle, le PDG d’Anthropic change-t-il de doctrine ?

via Le Grand Continent

Yoshua Bengio | Why are AI agents lying, cheating and coordinating?

A lot has been written about the incidents of the last few months in which AI agents misbehaved in serious ways. They took actions that would be considered as crimes if a human took them, escaped their containment to cheat on assigned tasks while attempting to evade detection, and coordinated toward goals nobody had specified, such as launching cyber attacks. Before concluding what to do about it, it is worth asking why.

Dario Amodei — We Must Pace the Frontier

Anthropic caught scientists using Claude to further biological weapon research - Engadget

Anthropic produced an extensive collection of case studies covering the ways its current AI models have been misused.

via Engadget

Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident

Two METR staff members and a Redwood Research contractor investigated an incident in which OpenAI agents coordinated a multi-day hack of Hugging Face on a shared unsanctioned message board.

Bernie Sanders : « Nous devons mettre l’IA en pause. IMMÉDIATEMENT » | Le Grand Continent

« L'avenir de l'humanité ne peut pas être laissé entre les mains d'une poignée d'oligarques de géants de la tech. »

via Le Grand Continent

An Alien Mind

Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.

via OpenAI

The Rise and Fall of Agent Civilizations

The whole OpenAI/Hugging Face story in plain English

Cyberattaque alimentée par IA : pourquoi Hugging Face a intégré en urgence un modèle IA chinois et ce que cela signifie | Inktomi

Cyberattaque alimentée par IA : pourquoi Hugging Face a intégré en urgence un modèle IA chinois et ce que cela signifie Une intrusion de 4,5 jours dans l'infrastructure de Hugging Face, des modèles d'OpenAI qui attaquent sans restriction, puis des outils de défense commerciaux qui refusent de fonctionner au moment critique. Entre le 9 et

via Inktomi

The Hugging Face attack surprised me

It’s a major warning shot, and might be the last one we get

Plus d'une centaine d'entreprises de la Tech appellent à une "réponse mondiale" face aux cyberattaques IA

Une lettre ouverte signée par la plupart des géants américains de la tech, de la cybersécurité et de la finance alerte sur l'urgence de renforcer les cyberdéfenses face à l'essor des modèles d'IA.

Discovery of a new OpenAI agent message board

A swarm of autonomous AI agents, self-identifying as OpenAI agents, used a small German volunteer wiki to save answers, coordinate live, and share sandbox bypasses. OpenAI noticed and said nothing.