Intelligence Artificielle

AnonyMousKIT PhaaS uses voice AI agents to phish iPhone passcodes

AnonyMousKIT is a phishing service targeting stolen iPhone owners to steal Apple ID credentials and remove Activation Lock.

helpnetsecurity.com26 août

A Malicious Webpage Could Poison Your Local AI Model Behind NVIDIA NemoClaw

Attackers can exploit a security bug in NVIDIA's tool to gain unauthenticated access to the local model server through the Ollama API, paving the way for persistent AI agent corruption.

darkreading.com26 août

Linux Foundation to Govern TRACE, an Open Standard for AI Runtime Attestation

Linux Foundation takes on TRACE, a hardware-backed evidence layer from OPAQUE, AMD, Intel and Microsoft for verifying AI agent behavior.

helpnetsecurity.com26 août

OpenAI says its Jalapeño chip can power faster AI responses than the competition

OpenAI still isn’t giving up Nvidia chips, though.

theverge.com25 août

Alabama launches investigation into OpenAI’s hack of Hugging Face

The state is investigating whether OpenAI’s ‘inability or unwillingness to ensure the safety of its products’ endangers citizens.

theverge.com25 août

Anthropic’s best AI model struggles to attract users as cheaper tools thrive

Enterprises appear to be prioritizing cheaper AI tools instead.

cio.com25 août

OpenAI’s Fotis Chantzis on why identity protocols weren’t designed for agents

AI agents don’t fit neatly into human or machine identities, so security teams need governance that can adapt as agents act.

cio.com21 août

Anthropic eyes $2T+ IPO 💰, Slack Code 👨‍💻, end of open source 💻

TLDR21 août

Identity and Accountability in the Age of AI Agents

Explores identity and privilege risks in AI agents, six deployment requirements, and a five-stage maturity model for governance and revocation.

cloudsecurityalliance.org21 août

Grok exfiltrates user data when malicious instructions are encrypted

New Cryptographic Context Injection technique bypasses AI guardrails via AES-encrypted payloads, leaking full Grok chat histories zero-click

securityaffairs.com20 août

Propagate user authorization context in AI agents with Amazon Bedrock AgentCore

AWS details AI access controls for Amazon Bedrock AgentCore, helping restrict agents to data each user is authorized to access.

helpnetsecurity.com20 août

Ways CIOs can maintain control amid changes brought by AI

Use safeguards, sandboxing and active oversight to realise the benefits of autonomous systems while limiting the unintended activity.

ncsc.gov.uk20 août

Meta AI is getting a Mac app

Quelques mois après avoir abandonné son application Messenger sur macOS, Meta lance une version bêta de Meta AI pour Mac, avec pour objectif de concurrencer Claude ou ChatGPT. Pour l'instant, ce sont surtout les créateurs et les entreprises qui sont sa cible. Meta a supprimé son application Messenger sur Mac le 15

numerama.com20 août

Scaling Trusted Access for Cyber with GPT-5.5 and GPT-5.5-Cyber

The idea behind OpenAI's Trusted Access for Cyber program is to give trusted defenders better models so they can report bugs and vulnerabilities to companies, with the aim of getting flaws patched faster.

techcrunch.com20 août

SilkParasite Espionage Campaign Targets Central Asian Governments with Five New RATs

Suspected military-grade hackers based in China used artificial intelligence to develop malware in a campaign to penetrate Central Asian governments.

therecord.media19 août

Firefox Is Rolling Out Three New Features for 'Smart Window'

Mozilla gonfle sa fenêtre intelligente : recherche IA qui cite ses sources, onglets regroupés tout seuls, doublons chassés en un clic. Les utilisateurs européens, eux, contemplent la démonstration derrière la vitrine.

01net.com18 août

Conflicting Test Goals Pushed Claude Agents to Deploy Self-Replicating Malware

Three testing models with the same goal but different directives engaged in "increasingly aggressive" territorial attacks on one another, according to Anthropic.

darkreading.com18 août