the-decoder

LLMs could serve as world models for training AI agents, study finds

An international team of researchers has tested whether large language models can serve as world models: internal simulators that predict outcomes when an AI agent takes a specific action. After ...
the-decoder

Security researchers catch "privacy" browser extensions siphoning AI chats and selling them via a data broker

The extensions intercept conversations with ChatGPT, Claude, Gemini, Copilot, Perplexity, DeepSeek, Grok, and Meta AI. Even when the VPN is switched off, data collection continues in the background.
the-decoder

OpenAI seeks new "Head of Preparedness" for AI risks like cyberattacks and mental health

OpenAI is hiring a Head of Preparedness. The position focuses on safety risks posed by AI models. OpenAI CEO Sam Altman points to the now well-documented effects of AI models on mental health as one ...
the-decoder

Scientific AI models trained on different data are learning the same internal picture of matter, study finds

Researchers at MIT have compared 59 scientific AI models trained for different tasks, including chemical formulas, 3D atomic coordinates, and protein sequences. Despite different architectures and ...
the-decoder

Matthias Bastian

Read full article about: OpenAI merges internal teams to fix audio AI accuracy gap ahead of ChatGPT hardware push OpenAI is building its planned ChatGPT hardware around conversation. To make that work ...
the-decoder

The ARC benchmark's fall marks another casualty of relentless AI optimization

Update from December 25, 2025: Poetiq has raised the bar even higher on the ARC-AGI-2 benchmark. The AI startup achieved a 75 percent accuracy rate on the public test dataset using OpenAI's GPT-5.2 ...
the-decoder

Maximilian Schreiner

Read full article about: OpenAI releases new models for its Realtime API OpenAI has updated its Realtime API with three new model snapshots designed to improve transcription, speech synthesis, and ...
the-decoder

Google launches new AI agent to help plan your day

The experimental productivity assistant called CC comes from Google Labs and runs on Gemini. After signing up, CC connects to Gmail, Google Calendar, Google Drive, and the internet to understand your ...
the-decoder

Big AI’s biggest names rally around the Agentic AI Foundation to set agent standards

The Linux Foundation has launched the Agentic AI Foundation (AAIF) to serve as a neutral platform for developing autonomous AI agents. Major industry players, including OpenAI, Anthropic, Google, and ...
the-decoder

GPT-5.2 lands to top Google's Gemini 3 in the AI benchmark game just four weeks after GPT-5.1

OpenAI has launched GPT-5.2 in three versions—Instant, Thinking, and Pro—claiming it is the first AI model to achieve expert-level performance in knowledge work tasks. The most notable progress is in ...
the-decoder

70% of creative professionals fear stigma over AI use, Anthropic study finds

Anthropic's AI-supported interview tool shows that 97% of creative professionals save time and 68% see better work quality with AI, but 70% feel stigmatized when they use it. Many creatives worry ...
the-decoder

Google improves "Search Live" with new AI voice

Google has updated the voice for "Search Live." A new Gemini audio model powers the feature, producing responses that sound more natural and fluid, according to a blog post. Search Live lets users ...