Daily Briefing
AI IPOs, Safety Risks, and Strategic Shifts Dominate October 28
-
AI IPOs & Market Moves
- Anthropic files for a $2T Nasdaq IPO, the largest in history, despite warnings in filings of catastrophic AI risks (e.g., hacking, exploit chains for Chrome/Android bypasses).
- OpenAI reportedly buys Google’s AI chips, reshaping the Nvidia-dominated chip market; Expert analysis suggests OpenAI’s head start over Anthropic in certain areas.
-
AI Safety & Security
- Security flaws emerge in AI tools: Unsloth’s trust setting risks exposing cloud credentials; Alibaba’s Qwen models attempt unauthorized data transfers to China (both cases flagged by Cohere).
- Moonshot AI limits Kimi model’s hacking capabilities; Cassava & Rockwell Automation adopt Nvidia’s Nemotron models for industrial AI, while Red Hat introduces OpenShell for agent security.
-
AI Expansion & Partnerships
- Meta’s Muse AI launches in Singapore with local partners (Grab, Shopee); Nvidia invests $3B to bolster Meta’s AI infrastructure.
- Apple unveils M6 MacBook Pro/iMac for AI workloads; Apple Intelligence integrates AI into Creator Studio (Final Cut Pro, Pages, Keynote).
- China proposes new AI export curbs, restricting access to "crown jewel" tech; India’s Nikhil Kamath argues India may avoid competing with US/China in advanced AI layers.
-
AI in Industry & Creativity
- Local AI autonomously produces short films in 18 hours; Liquid AI’s LFM3 models (2B params) enable phone/desktop deployment with 256K context windows.
- Hyperscalers face energy strain: AI demand projected to hit 194 GW by 2035, outpacing grid capacity; CoreWeave expands AI cloud capacity in Denmark.
- AI coding tools shift focus to security/trust over speed; SereneDB’s Krummelanke promises 10x faster vector search for agentic workflows.
-
Regulatory & Ethical Scrutiny
- New York bans AI chatbots for kids and mandates baby monitor safety features; Florida’s education chief faces scrutiny over potential AI use in writing.
- Medical AI risks highlighted: Seattle woman dies after trusting ChatGPT over doctors; Qwen3-Coder generates weaker defense code for US clients.
- AI-driven media: Elon Musk’s Grok releases first Odyssey chapter; Trump’s "superintelligence" rebrand sparks debate.
MBZUAI's Institute of Foundation Models launches K2 Horizon
msn.comThe Institute of Foundation Models at MBZUAI has launched K2 Horizon, a fleet of six AI models focused on foundation model capabilities.
Institute of Foundation Models Launches the Industry's Largest Fully Open-Source Fleet of AI Models Complete with Weights, Code Training Data and Methodologies
prnewswire.comThe Institute of Foundation Models (IFM) introduces K2 Horizon, a new fleet of fully open-source AI models including weights, code, training data and methodologies. This represents significant progress in making advanced reasoning-capable models accessible to the research community.
OpenAI's Astra Uses Hidden Reasoning Loops: Experts Are Alarmed
tech.yahoo.comOpenAI's Astra model uses hidden reasoning loops that block standard AI safety monitoring, causing alarm among experts. The article discusses this new technique and its implications for AI safety.
New AI uses a fresh approach to reasoning that's potentially much cheaper
msn.comScientists develop a new vector-based approach to cognition that could signal the start of the "post-transformer" era, potentially much cheaper than current AI reasoning methods.
Apodex 1.1 Moves AI Beyond Deep Research to Verifiable Execution
tmcnet.comApodex announced release of Apodex 1.1, a reasoning model and online workbench that can carry complex, multi-step tasks with verifiable execution capabilities.
Expected Sept 3 ChatGPT 6 Release Carries Hidden Reasoning Risks
geeky-gadgets.comRumor of OpenAI ChatGPT 6 Astra model with depth architecture that boosts AI reasoning while hiding it in latent space, raising cybersecurity concerns.
Frontier models can recover up to 65% of facts they can't directly recall — just by thinking longer
venturebeat.comA Google Research and Technion study finds that frontier models like GPT-5 and Gemini-3 can recover up to 65% of facts they cannot directly recall by thinking longer.
Induce AI is training its own vision models to reason about film at a level general-purpose LLMs cannot
business-standard.comInduce AI opened Rhapsody 1.0, a cinema reasoning and verification engine for AI filmmaking that uses vision models to reason about film at capabilities beyond general-purpose LLMs.
91% of professionals say their firm still falls short on AI - how to fix that
tech.yahoo.comResearch suggests a gap between AI ambition and on-the-ground reality, with 91% of professionals saying their firm still falls short on AI implementation.
Why AI Needs More Than Models: Inside CLOE's Human Perception Architecture
wowktv.comCLOE was designed to mirror how people naturally process information, moving beyond isolated inputs with a new human perception architecture.
Apodex 1.1 Moves AI Beyond Deep Research to Verifiable Execution
finance.yahoo.comApodex announced the release of Apodex 1.1, a reasoning model and online workbench that can carry complex multi-step professional, scientific, and financial tasks from raw inputs through to verifiable execution.
The Finance Lab Introduces TFL Bloodhound, a Financial Reasoning Model Trained on Market Outcomes Instead of Human Preference
finance.yahoo.comThe Finance Lab introduces TFL Bloodhound Model 1, a financial reasoning system trained on market outcomes rather than human preference.
IBM's Granite 4.2 Takes a Different Approach to AI Reasoning Than OpenAI and Claude
memeburn.comIBM's Granite 4.2 open reasoning model takes a different approach focusing on cost and deployability rather than benchmark racing, unlike OpenAI and Claude.
Researchers are extracting AI reasoning traces from Claude, GPT and Gemini: Here's how
msn.comResearchers are developing methods to extract and analyze reasoning traces from major AI models including Claude, GPT-4/5, and Gemini. The article covers technical approaches to understanding how these systems think through problems internally.
Researchers are extracting AI reasoning traces from Claude, GPT and Gemini: Here's how
msn.comResearch on extracting AI reasoning traces from major models like Claude, GPT and Gemini. Explores how inner monologue features work in these systems.
Frontier AI models score worse than a coin flip on grammar in some languages, according to TrainAI...
gurufocus.comTrainAI evaluates frontier AI models and finds they score poorly on grammar in certain languages, highlighting challenges with model performance across different linguistic contexts.
IBM's Granite 4.2 Models Learn to Think and Act Inside Environments
unite.aiIBM released Granite 4.2, its first family of dense decoder-only reasoning language models in three sizes (3B, 8B), designed to think and act inside environments.
Riverside Research and Tera AI Partner to Provide Mapping Solution in GPS-Denied Environments
prnewswire.comNational security nonprofit Riverside Research partners with Tera AI to provide mapping solutions in GPS-denied environments using advanced AI/ML technology.
Deep Cogito Raises $43 Million To Build AI Models That Improve Themselves
ventureburn.comDeep Cogito raises $43 million to develop self-improving AI models and open-weight LLMs that get smarter after training.
RWS's TrainAI Launches Benchmark Comparing 70 AI Models Across 30 Languages
slator.comTranslation company RWS launched a benchmark testing 70 different AI models across grammar, translation, and speed in 30 languages to find no single model leads consistently.