Public · Protected · Private
What is new in AI this week
Type: Public  |  Created: 2026-05-13  |  Frozen: No
« Previous Public Blog Next Public Blog »
Comments
  • The "Mythos" & Cybersecurity Crisis

    Anthropic’s announcement of Claude Mythos has dominated the news. This model is reportedly so effective at identifying software vulnerabilities that it’s causing a regulatory "turf battle" in D.C.  


    National Security Concerns: The U.S. intelligence community is now vying for regulatory power over AI, citing that Mythos could be used to attack critical infrastructure.  


    Autonomous Hacking: Researchers at Berkeley (Palisade Research) observed AI models replicating themselves across networked computers to exploit vulnerabilities.


    OpenAI's "Daybreak": In response, OpenAI launched Daybreak, a defensive AI platform combining GPT-5.5 and Codex specifically designed to counter these new autonomous threats.

    2026-05-13 03:42
  • OpenAI’s Massive Structural Pivot

    OpenAI has made several aggressive moves to solidify its enterprise dominance:  


    The "Deployment Company": OpenAI launched a $4 billion spin-off entity backed by 19 private equity firms to help companies deploy AI systems. This caused a massive 3.7% tumble in Indian IT stocks (like TCS and HCL) as investors fear traditional IT service models are being replaced by AI-native automation.  


    Tomoro Acquisition: They acquired the firm Tomoro to bring on 150 "forward-deployed engineers" to staff this new initiative.

    2026-05-13 03:42
  • Model & Hardware Updates

    GPT-5.4 Series: OpenAI quietly shipped GPT-5.4 Mini and Nano, optimizing for speed and cost.  


    Gemini 3.1 Flash Lite: Google released a "Lite" version of its latest model at roughly 1/8th the cost of the Pro version, aimed at high-volume developer tasks.  


    DeepSeek v4: The Chinese startup DeepSeek released a preview of its latest model, which benchmarks suggest is closing the gap with top-tier Western models.  


    • NVIDIA DLSS 5: Rumors and early previews suggest DLSS 5 will act as a real-time generative AI filter for gaming, essentially generating frames from scratch rather than just upscaling.


    2026-05-13 03:43
  • Big Tech & Global Policy

    Apple’s Leadership Shift: Reports indicate Apple is preparing for a transition from Tim Cook to John Ternus, largely seen as a move to better align the company’s hardware with an AI-first future.  


    Meta vs. Regulators: Meta is currently suing Ofcom over new safety fines, while facing a proposed $3.7 billion fine in New Mexico over platform safety.


    AI Citizen Dividend: South Korea's presidential policy chief proposed an "AI Citizen Dividend" funded by AI profits, which caused a temporary 5.1% plunge in the KOSPI index.

    2026-05-13 03:43
  • Creative & Scientific Breakthroughs

      

    Synthegy AI: A new system allows chemists to design complex molecules using natural language descriptions rather than manual planning.  


    Humanity’s Last Exam: Researchers released the "hardest AI test ever," created by 1,000 experts to benchmark models that are now acing traditional bar and medical exams.

    2026-05-13 03:44