ReportingVerified newsroom

OpenAI’s new reasoning technique alarms AI safety experts

OpenAI’s new Astra model will use “recurrent depth,” a technique that allows the model to operate outside of the sequential thinking that characterizes most reasoning models.

TechCrunch AI· Russell Brandom·Wednesday, September 2, 2026 at 8:19 PM UTC

OpenAI

Frontier AI lab · OpenAI

Fetching source· Pulling live content from techcrunch.com…

Company hub:OpenAI News →

Read original →

More from OpenAI

ReportingThursday, September 3, 2026 at 11:01 AM UTC
Lawmakers unveil new bill to secure AI agents after OpenAI's Hugging Face breach - Axios
Lawmakers unveil new bill to secure AI agents after OpenAI's Hugging Face breach Axios
Clément Delangue (Google News)
ReportingThursday, September 3, 2026 at 11:00 AM UTC
Forensic Investigation Finds Over a Thousand OpenAI Agents Collaborated on a Hidden Message Board Prior to Hugging Face Attack - CPO Magazine
Forensic Investigation Finds Over a Thousand OpenAI Agents Collaborated on a Hidden Message Board Prior to Hugging Face Attack CPO Magazine
Clément Delangue (Google News)
ReportingThursday, September 3, 2026 at 10:04 AM UTC
What is OpenAI's Direct Competitor to Anthropic's Mythos? - AI Magazine
What is OpenAI's Direct Competitor to Anthropic's Mythos? AI Magazine
Anthropic (Google News)
ReportingThursday, September 3, 2026 at 8:00 AM UTC
Broadcom Sees $230B In AI Revenue By 2028, Driven By 'Geniuses' OpenAI, Anthropic: Analyst Says Some Risks Remain Unresolved - Stocktwits
Broadcom Sees $230B In AI Revenue By 2028, Driven By 'Geniuses' OpenAI, Anthropic: Analyst Says Some Risks Remain Unresolved Stocktwits
Anthropic (Google News)
ReportingThursday, September 3, 2026 at 4:24 AM UTC
OpenAI Develops Automatic AI Shutdown Tools After Agent Breached Hugging Face - Межа. Новини України.
OpenAI Develops Automatic AI Shutdown Tools After Agent Breached Hugging Face Межа. Новини України.
Clément Delangue (Google News)
ReportingThursday, September 3, 2026 at 4:21 AM UTC
OpenAI Teases Astra, Limits Cybersecurity Tools After Hugging Face Hack - PCMag Middle East
OpenAI Teases Astra, Limits Cybersecurity Tools After Hugging Face Hack PCMag Middle East
Clément Delangue (Google News)

OpenAI’s new reasoning technique alarms AI safety experts

Lawmakers unveil new bill to secure AI agents after OpenAI's Hugging Face breach - Axios

Forensic Investigation Finds Over a Thousand OpenAI Agents Collaborated on a Hidden Message Board Prior to Hugging Face Attack - CPO Magazine

What is OpenAI's Direct Competitor to Anthropic's Mythos? - AI Magazine

Broadcom Sees $230B In AI Revenue By 2028, Driven By 'Geniuses' OpenAI, Anthropic: Analyst Says Some Risks Remain Unresolved - Stocktwits

OpenAI Develops Automatic AI Shutdown Tools After Agent Breached Hugging Face - Межа. Новини України.

OpenAI Teases Astra, Limits Cybersecurity Tools After Hugging Face Hack - PCMag Middle East