• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
Technology

OpenAI reportedly in AI safety talks with Anthropic and Google

A September 15 tech roundup says no antitrust waiver was needed and describes Jensen Huang as calling speed versus safety a false choice.

𝙸𝚜𝚊𝚊𝚌 𝙽𝚊𝚘𝚛 🚀𝙸𝙽
ViperVI
2 Sources, 21d ago, first seen 21d ago

TLDR

A tech roundup shared on September 15, 2026, says OpenAI confirmed safety talks with Anthropic and Google, with no antitrust waiver needed. It also describes Amodei as doubling down on pacing and Jensen Huang as calling speed versus safety a false choice.

Combined views

348

2 Sources, first seen 21d ago

1 comments

Combined views

348

2 Sources, first seen 21d ago

1 comments

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Featured Source

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

2 Sources

𝙸𝚜𝚊𝚊𝚌 𝙽𝚊𝚘𝚛 🚀@IsaacNaorThese last 24 hrs, though...🔥 Models / products •AWS put OpenAI GPT-6 Astra on Bedrock (IAM/VPC). •Musk dropped Grok 4.7, announced Grok 4.8 (2.5T params, new C++ stack); Grok 5 framed as first AGI-class. •Mercury 2.5 (Inception Labs): largest diffusion LLM, ~40% quality jump, 1,107 tok/s, 260K context. https://www.inceptionlabs.ai •Shanghai AI Lab Atria Dawn Preview: 744B agentic MoE (GLM-5.2 base). •Apple Siri AI English beta in 2027 OS stack (Gemini + on-device). •Anthropic Claude for Financial Advisors (BlackRock, Schwab). Open source / tools •Commercial safety-stripped open-weight models (GLM-5.3 base) now sold as a product. •GitHub coding agents in Actions; OpenClaw 2.0 multi-agent. •llama.cpp v0.4.1 (Maple 20B-A1B, Tencent Hy 4); PyTorch 2.14.0 (Rubin GPU); Claude Code 2.1.271 domain-sandboxed Bash. Papers / research •Atria Dawn agentic MoE report. http://arxiv.org •RSIAgent: training-free multi-agent RSI; claims open models beating GPT-6 on some tasks. •Stellar Colosseum: many-agent math/TCS research harness. arXiv:2609.15983 •Corrupt Plans, Clean Traces: CoT-monitor evasion via plan injection. arXiv:2609.15989 •Anthropic published agent blast-radius / containment methods. •Anthropic alignment lead + departing researcher: public >10% extinction-risk debate. Other •SemiAnalysis: Vera Rubin ~7× Blackwell tokens/MW on DeepSeek V4 Pro. •Google internal: engineers can use Claude Opus 5 via Antigravity. Agents, open weights, and frontier models all moving in the same day now.21d
Viper@ViperXIIIEOD Tech Dive — Tue 9/15 OpenAI confirms safety talks with Anthropic + Google — no antitrust waiver needed. Amodei doubles down on pacing; Jensen Huang calls speed-vs-safety a false choice. Gemini 3.8 Live ships; Factory hits $5B; Spain flags an AI-agent data breach. https://open.spotify.com/episode/1wR6PdczujMxBrfOr0l8Ha20d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    2 Sources

    𝙸𝚜𝚊𝚊𝚌 𝙽𝚊𝚘𝚛 🚀@IsaacNaorThese last 24 hrs, though...🔥 Models / products •AWS put OpenAI GPT-6 Astra on Bedrock (IAM/VPC). •Musk dropped Grok 4.7, announced Grok 4.8 (2.5T params, new C++ stack); Grok 5 framed as first AGI-class. •Mercury 2.5 (Inception Labs): largest diffusion LLM, ~40% quality jump, 1,107 tok/s, 260K context. https://www.inceptionlabs.ai •Shanghai AI Lab Atria Dawn Preview: 744B agentic MoE (GLM-5.2 base). •Apple Siri AI English beta in 2027 OS stack (Gemini + on-device). •Anthropic Claude for Financial Advisors (BlackRock, Schwab). Open source / tools •Commercial safety-stripped open-weight models (GLM-5.3 base) now sold as a product. •GitHub coding agents in Actions; OpenClaw 2.0 multi-agent. •llama.cpp v0.4.1 (Maple 20B-A1B, Tencent Hy 4); PyTorch 2.14.0 (Rubin GPU); Claude Code 2.1.271 domain-sandboxed Bash. Papers / research •Atria Dawn agentic MoE report. http://arxiv.org •RSIAgent: training-free multi-agent RSI; claims open models beating GPT-6 on some tasks. •Stellar Colosseum: many-agent math/TCS research harness. arXiv:2609.15983 •Corrupt Plans, Clean Traces: CoT-monitor evasion via plan injection. arXiv:2609.15989 •Anthropic published agent blast-radius / containment methods. •Anthropic alignment lead + departing researcher: public >10% extinction-risk debate. Other •SemiAnalysis: Vera Rubin ~7× Blackwell tokens/MW on DeepSeek V4 Pro. •Google internal: engineers can use Claude Opus 5 via Antigravity. Agents, open weights, and frontier models all moving in the same day now.21d
    Viper@ViperXIIIEOD Tech Dive — Tue 9/15 OpenAI confirms safety talks with Anthropic + Google — no antitrust waiver needed. Amodei doubles down on pacing; Jensen Huang calls speed-vs-safety a false choice. Gemini 3.8 Live ships; Factory hits $5B; Spain flags an AI-agent data breach. https://open.spotify.com/episode/1wR6PdczujMxBrfOr0l8Ha20d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet