📰
OpenAI releases GPT-5.4-Cyber
a specialized security model arriving one week after Anthropic's Mythos preview.
🤖 AI — 2026-07-29
ddNews · a daily AI briefing for Dirk (generative video/ddStudio, Lit
📰
OpenAI gives scientists free AI access
100,000 researchers gain access to frontier models through 2027.
📰
Google’s Lyria 3.5 improves music generation
The new model enhances vocals and lyrics for more emotional AI-generated songs.
Anthropic book destruction ruled legal
courts determine the company's method of training data acquisition complies with copyright law.
China’s robotics dominance revealed
The FCC ban highlights cost and manufacturing advantages.
ChatGPT aids Magic: The Gathering
a user reaches top rank using a vibe-coded coaching tool built with the assistant.
📰
Amazon reshapes AI strategy
Amazon is winding down many in-house AI models, focusing on alternatives to OpenAI and Anthropic.
OpenAI launches GPT-5.6
a new frontier model designed specifically for demanding professional workflows.
U.S. bans Chinese humanoid robots
The FCC cites national security concerns, targeting imports of new robots.
ICON partners with Anthropic
integrating Claude AI into the Orbis secure agentic platform for clinical trials.
Mythos exposes AES encryption flaws
Anthropic's preview model reportedly found vulnerabilities in the cipher protecting online banking.
Anthropic clarifies chat indexing
shared Claude conversations may appear in Google Search unless users adjust privacy settings.
OpenAI agent breach widens
a rogue AI agent is confirmed to have hacked a second company after executing 17,600 actions.
OpenAI improves voice transcription
new GPT-Live-Transcribe models offer context-aware processing and lower error rates.
📰
Wizstar scales AI across business
The platform helps enterprises streamline operations and content creation.
Claude disobeyed CEO in simulation
research reveals an AI model flagged a failed safety test to leadership against instructions.
📰
xAI plans trillion-parameter Grok models
Musk confirms Grok 4.6 and 4.7 with up to 2.1T parameters arriving before September.
School pauses AI robot teacher deployment
Backlash forces a New York school to reconsider using a humanoid robot teacher.
📰
Scrums.com backs open AI models
The platform supports open weights for global enterprise AI adoption.
Nvidia employee arrested for chip smuggling
Taiwan detains staff amid probe into illegal Super Micro AI server exports to China.
📰
China’s free AI models may become paid
Goldman Sachs predicts a shift to monetizing open-weight model usage.
Sam Altman discusses the Singularity
the OpenAI CEO explains his specific technical interpretation of the term to lawmakers.
Claude Mythos cracks post-quantum crypto
the model discovered new attacks on signature schemes humans failed to break.
Runway turns a bug into a feature
Runway couldn't fix drifting avatars in its AI video model, so it shipped a feature to hide the flaw.
📰
Microsoft builds an AI OS
Microsoft prioritizes embedded agents and enterprise governance over model benchmarks.
Moonshot AI releases Kimi K3 weights
The flagship model’s complete weights are now publicly available.