Google Unveils Gemini 4 Argon, Limited to Trusted Cyber Defenders First

Today's 72–7

1. industry · Ars Technica AI · 2건

   OpenAI Scraps Model Over Safety, Delays IPO

   OpenAI delays IPO over AI safety concerns

   OpenAI reportedly ditched a model after a top executive told the Wall Street Journal it showed poor aptitude for following orders. Separately, the company's IPO plans have slipped amid AI safety concerns, and it is seeking another $30 billion privately.

   This signals how instruction-following failures can block a model's release, a risk factor developers should weigh when choosing which models to deploy in production.

   Ars Technica AI·TechCrunch AI
2. industry · TechCrunch AI · 2건

   OpenAI in Talks for $30B Round at $1.4T Valuation

   OpenAI reportedly in talks to raise $30B round at $1.4T valuation

   OpenAI is reportedly in talks to raise a $30B round at a $1.4T valuation. The round is anticipated to be the company's last before its delayed 2027 public debut.

   This mega-round shapes OpenAI's war chest for research and infrastructure, which can directly affect the pricing and roadmap of models and APIs developers rely on.

   TechCrunch AI·TechCrunch AI 2
3. research · Hacker News

   GLM-5.3 Raises Concerns Over Spread of Advanced Cyber Capabilities

   GLM-5.3 and the spread of advanced cyber capabilities

   GLM-5.3 is being discussed in connection with the spread of advanced cyber capabilities. The topic was shared and discussed on Hacker News.

   As powerful open models raise risks of misuse for building attack tools, teams need to revisit deployment and access control policies.

   Hacker News 100점
4. research · Google DeepMind

   DeepMind Unveils SynthID Bio to Watermark AI-Generated Proteins

   Introducing SynthID Bio

   Google DeepMind introduced SynthID Bio, a proof of concept for watermarking AI-generated proteins. The technique preserves the biological function of proteins while embedding the watermark.

   This gives developers working on AI-designed biomolecules a new way to verify provenance and guard against misuse in their pipelines.

   Google DeepMind
5. policy · OpenAI

   OpenAI Disrupts Coordinated Model-Distillation Campaign

   Disrupting a coordinated model-distillation campaign

   OpenAI disrupted a coordinated campaign attempting to extract its protected model reasoning through distillation. The company is strengthening defenses against such adversarial distillation attempts.

   Teams pulling large volumes of model outputs via API to train competing models should review terms-of-service risk and usage boundaries.

   OpenAI
6. release · Simon Willison · 2건Korean coverage · THE AI

   Claude Sonnet 5.5

   THE AI·Simon Willison