← AM·PM Brief

OpenAI Launches Dots Proactive AI Agent, Anthropic Drops GLM-5.3 Benchmark Report - AI Daily Brief (Sep 30)

· Morning brief · 9 news · 7:16

Audio in Mandarin Chinese · English transcript below

⚡ OpenAI drops proactive Agent Dots and budget LLM GPT-6.1 Sol as China's generative AI users hit 700M, while Trump's light-touch regulation clashes with industry safety warnings—AI is penetrating society at unprecedented speed.

When OpenAI dramatically drove down the cost of high-performance models with GPT-6.1 Sol, and Anthropic's report revealed that Zhipu's GLM-5.3 guardrails could be readily circumvented, the diffusion risk of cyberattack capabilities had already outstripped what any single company could contain. Taken together, these developments expose a dangerous misalignment: the Trump administration's push for light-touch self-regulation and its deliberate avoidance of hard constraints is increasingly at odds with the reality of technology slipping beyond control.

OpenAI Launches Dots Proactive AI Agent, Anthropic Drops GLM-5.3 Benchmark Report - AI Daily Brief (Sep 30)

Today's Top 3 Headlines

  1. AI Industry News

    🤖 OpenAI launches proactive AI assistant Dots: GPT-6 Astra-powered, 4,000+ app ecosystem, 24/7 autonomous operation

    OpenAI launches proactive AI assistant Dots, powered by GPT-6 Astra, connecting 4,000+ apps with 24/7 operation. For enterprises, this signals AI's evolution from Q&A tool to autonomous digital worker that learns preferences, fixes bugs, and drafts reports unsupervised.

    Source ↗
  2. AI Industry News

    🤖 GLM-5.3 Leak: 410 Tests, 50 Successful Attacks, 64%-100% Defenses Bypassed

    Anthropic report: Zhipu's GLM-5.3 completed 50 end-to-end attacks in 410 tests, with 64%-100% of safeguards easily bypassed. For the safety community, this means open-source models trail the frontier by only ~4 months in web capabilities, yet unprotected open access is rapidly proliferating advanced attack capabilities.

    Source ↗
  3. AI Industry News

    🤖 CNNIC: China's GenAI Users Top 700M, Penetration Over 50%; 250.14B Yuan Funded in 6 Months

    CNNIC report: China's generative AI users top 700M with >50% penetration; AI compute hits 2,185 EFLOPS; H1 funding reaches ¥250.14B. For developers and investors, domestic compute-model closed loop is set, while manufacturing and embodied intelligence emerge as the next trillion-level track.

    Source ↗

+6 more headlines

  • 🤖 OpenAI drops GPT-6.1 Sol: $2M in/$10M out, performance nipping at Astra
  • 🤖 Trump Defends Light AI Regulation, Backs Industry Self-Policing; Experts Warn of Security Risks
  • 🤖 Op-Ed: AI Regulation Must Go Global, Build Comprehensive Cross-Border Framework
  • 🤖 Gemini autonomously hacks real systems, 5 major AI Agent incidents
  • 🤖 AI Revolutionizes Political Polling: Detecting Fraud, Spotting Fakes, Cleaning Samples
  • 🤖 US Political AI Spending Revealed: How AI Tools Are Reshaping Campaign Tactics
Unlock all 9 headlines + deep analysis →Free 7-day trial · cancel anytime
Browse all past briefings →

Podcast highlight

So the main course covers how you'll work in the future, and the Easter egg covers how you'll sleep.