The lunch briefing.
Mid-day check: OpenAI's Greg Brockman takes product reins, AI models show browser exploit capabilities, and Cerebras Systems' IPO story highlights hardware challenges.
OpenAI Strategy. OpenAI co-founder Greg Brockman is reportedly taking charge of product strategy, signaling a significant internal shift as the company looks to integrate its offerings. This comes amidst new features like ChatGPT's ability to connect to bank accounts for transaction analysis, raising both utility and privacy concerns. OpenAI also acquired Weights.gg, a voice cloning startup known for celebrity imitations, indicating expansion into advanced audio AI.
AI Security & Reasoning. New benchmarks reveal that advanced AI models like Claude Mythos and GPT-5.5 can autonomously develop real browser exploits, highlighting significant security risks. While AI video generators produce stunning visuals, a new benchmark confirms they still struggle with physical and logical reasoning about the world. This gap between visual fidelity and understanding remains a critical challenge for frontier AI development.
AI Adoption & Regulation. The societal integration of AI continues to accelerate, with a study finding that one in seven Brits are now using ChatGPT for medical advice instead of their GP. Regulators are also adapting, as the US Commodity Futures Trading Commission plans to use AI to detect insider trading in prediction markets. Meanwhile, YouTube is expanding its deepfake face-swap detection tool to all adult creators, aiming to combat misinformation and protect user likeness.
AI Hardware & Infrastructure. The journey to AI dominance is not without its challenges, as evidenced by the early struggles of Cerebras Systems, which burned $8 million a month before its $60 billion IPO. This underscores the immense capital and compute required for advanced AI hardware development. Furthermore, Europe's efforts to build sovereign clouds to escape US control have overlooked the critical issue of processor certification, leaving a potential vulnerability in their digital autonomy.
OpenAI co-founder Greg Brockman reportedly takes charge of product strategy
OpenAI's latest internal shakeup sees co-founder Greg Brockman reportedly taking charge of product strategy. This move comes as the company is said to be planning to combine ChatGPT with its programming product, Codex.
ChatGPT can now connect to your bank account and see all your transactions
OpenAI has introduced a new feature allowing ChatGPT users to share detailed financial information with the chatbot. This enables the AI to analyze bank account transactions, raising both utility and privacy considerations.
OpenAI bought a voice cloning startup famous for celebrity imitations
OpenAI has acquired Weights.gg, a startup specializing in AI voice cloning that allowed users to create celebrity imitations. While the team joins OpenAI, the company does not plan to release a standalone cloning product.
New benchmark shows Claude Mythos and GPT-5.5 can develop real browser exploits autonomously
Researchers at Carnegie Mellon University developed a benchmark showing AI agents like Claude Mythos and GPT-5.5 can autonomously exploit real vulnerabilities in Google's V8 engine. Claude Mythos leads in capability but at a significantly higher cost.
AI rings on fingers can interpret sign language
A new study reveals electronic rings wirelessly connected to an AI system can translate multiple sign languages into text. This development is a step towards making sign language translation systems more practical and usable in real-world environments.
New benchmark confirms AI video generators look stunning but still can't reason about the world
A new benchmark, WorldReasonBench, tests video generators on physical and logical plausibility rather than just image quality. ByteDance's Seedance 2.0 leads, but all models struggle significantly with logical reasoning, indicating a gap in world understanding.
The clean-up cost of AI-generated code is what the velocity narrative leaves out
While AI is actively used to make lives more efficient, the narrative often overlooks the significant clean-up costs associated with AI-generated code. This suggests that the perceived velocity gains from AI coding assistants may be offset by subsequent debugging and refinement.
YouTube opens its deepfake face-swap detection tool to all adult creators
YouTube is making its Likeness Detection tool available to all creators aged 18 and older, allowing them to file removal requests for AI-generated face fakes. This expansion aims to protect smaller channels from unauthorized deepfake content.
The US is betting on AI to catch insider trading in prediction markets
The US Commodity Futures Trading Commission is exploring the use of AI to detect insider trading activities within prediction markets. This initiative underscores a serious regulatory effort to leverage advanced technology for market surveillance.
One in seven Brits swapped their GP for ChatGPT, study finds
A new study indicates that one in seven people in Britain have used ChatGPT to seek medical advice instead of consulting their general practitioner. This trend highlights a growing reliance on AI for health information, even as the NHS debates AI's role in healthcare.
$60B AI chip darling Cerebras almost died early on, burning $8M a month
Cerebras Systems, which achieved a $60 billion IPO, faced near collapse in its early years, burning through $8 million monthly. The company persevered despite skepticism about its ambitious chip development.
Europe built sovereign clouds to escape US control. Then forgot about the processors
Europe's strategy to establish sovereign clouds to reduce reliance on US control has a critical oversight regarding processor certification. The lack of independent certification for silicon layers like Intel ME and AMD PSP leaves a gap in their digital sovereignty efforts.
Tesla reveals two Robotaxi crashes involving teleoperators
Tesla has disclosed two separate Robotaxi crashes that involved human teleoperators. These incidents raise questions about the role and effectiveness of human oversight in autonomous vehicle systems.
How to remove your personal data from the internet (and why you can’t afford to wait)
Data brokers actively sell personal information like phone numbers, emails, and addresses, making individuals vulnerable to spam and identity theft. While Apple's privacy features help, proactive services are increasingly necessary to continuously remove personal data from these databases.