Minerval
Sourceunknowntext stored

OpenAI's new reasoning technique alarms AI safety experts

Russell BrandomTechCrunchPublished Sep 2, 2026Original︎ ↗Archived copy

Claims2
ExaminationsNone
Text115 sections
VersionsOne known
CopiesNone recorded
CorrectionsNone recorded
Facts checked Nov 11, 2026
ReachNo copies recorded
StructureIn 2 claims' maps0 sources downstream across 2 claims
EvidenceNot yet read
ReviewsNot reviewed

Examinations

None recorded. An examination is commissioned by a claim’s Steward for what its claim rests on; a review of the document for its own sake is asked for as a request bearing on it.

Text18 passages · 1 marked

No examination yetMarked passage
TechCrunch Desktop Logo0 passages
TechCrunch Mobile Logo0 passages
Latest0 passages
Startups0 passages
Venture0 passages
Apple0 passages
Security0 passages
AI0 passages
Apps0 passages
Disrupt 20260 passages
Events0 passages
Podcasts0 passages
Newsletters0 passages
Search Submit0 passages
Site Search Toggle0 passages
Mega Menu Toggle0 passages
Topics0 passages
Latest0 passages
AI0 passages
Amazon0 passages
Apps0 passages
Biotech & Health0 passages
Climate0 passages
Cloud Computing0 passages
Commerce0 passages
Crypto0 passages
Enterprise0 passages
EVs0 passages
Fintech0 passages
Fundraising0 passages
Gadgets0 passages
Gaming0 passages
Google0 passages
Government & Policy0 passages
Hardware0 passages
Instagram0 passages
Layoffs0 passages
Media & Entertainment0 passages
Meta0 passages
Microsoft0 passages
Privacy0 passages
Robotics0 passages
Security0 passages
Social0 passages
Space0 passages
Startups0 passages
TikTok0 passages
Transportation0 passages
Venture0 passages
More from TechCrunch0 passages
Staff0 passages
Events0 passages
Startup Battlefield0 passages
StrictlyVC0 passages
Newsletters0 passages
Podcasts0 passages
Videos0 passages
Partner Content0 passages
TechCrunch Brand Studio0 passages
Contact Us0 passages
Image Credits: SeongJoon Cho/Bloomberg / Getty Images0 passages
AI0 passages
OpenAI’s new reasoning technique alarms AI safety experts0 passages
Russell Brandom0 passages
1:19 PM PDT · September 2, 202612 passages · 1 marked

“My biggest concern is that a natural progression from here would involve scaling up the opaque reasoning to the point where the model reasons entirely or almost entirely in latent space,” Greenblatt wrote. “I hope it isn’t too late to avoid the most concerning architectures and that OpenAI will stop here.”

Topics0 passages
AI , ai safety , OpenAI1 passage
Russell Brandom0 passages
AI Editor1 passage
View Bio0 passages
20270 passages
San Francisco0 passages
Get an Early Peek at the Next TechCrunch Disrupt1 passage
Sign up for our waitlist0 passages
Register Now0 passages
Most Popular0 passages
Jensen Huang’s alleged honeypot scandal has inspired some top-notch memes0 passages
Dominic-Madori Davis0 passages
Lucas Ropek0 passages
The race is on to sue Nvidia0 passages
Connie Loizos0 passages
Waymo robotaxis kept illegally passing school buses after software fix, feds say0 passages
Kirsten Korosec1 passage
Kirsten Korosec0 passages
Tesla will stop making the Cybertruck in 20270 passages
Sean O'Kane0 passages
Claude can now analyze your spending with its new Personal Finance tool0 passages
Aisha Malik0 passages
Harbinger raises $240M after expanding into defense0 passages
Sean O'Kane0 passages
Loading the next article0 passages
Error loading the next article0 passages
X0 passages
LinkedIn0 passages
Facebook0 passages
Instagram1 passage
Mastodon0 passages
Threads0 passages
Bluesky0 passages
TechCrunch0 passages
Staff0 passages
Contact Us0 passages
Advertise0 passages
Site Map0 passages
Terms of Service0 passages
Privacy Policy0 passages
RSS Terms of Use0 passages
Code of Conduct0 passages
Voice AI0 passages
Anthropic0 passages
Flock0 passages
Startup Battlefield0 passages
Disrupt 20260 passages
Tech Layoffs0 passages
ChatGPT1 passage

Not found in the text 3

As quoted

OpenAI has worked to preserve and utilize chain-of-thought monitoring since our very first reasoning models. It's a core goal of our current research program.

As quoted

My biggest concern is that a natural progression from here would involve scaling up the opaque reasoning to the point where the model reasons entirely or almost entirely in latent space. I hope it isn't too late to avoid the most concerning architectures and that OpenAI will stop here.

As quoted

OpenAI has worked to preserve and utilize chain-of-thought monitoring since our very first reasoning models. It's a core goal of our current research program.

Lineage across 2 claims

Draws on · none recorded
This source
OpenAI's new reasoning technique alarms AI safety experts
in 2 claims' maps
Drawn on by · none recorded

Record

Kind
unknown
Authors
Russell Brandom
Publisher
TechCrunch
Published
Sep 2, 2026
DOI
None
Link
https://techcrunch.com/2026/09/02/openais-new-reasoning-technique-alarms-ai-safety-experts/
Archived copy
https://web.archive.org/web/20261017224046/https://techcrunch.com/2026/09/02/openais-new-reasoning-technique-alarms-ai-safety-experts/
Stored text
6,174 characters, retrieved Nov 11, 2026 · 115 sections, 18 passages
Claims extracted
Not recorded
Versions
One known
Copies
None recorded
Corrections
None recorded
Facts checked
Nov 11, 2026

History 3

  1. Nov 11, 2026DividedSegmenter · Divided into 133 parts by its own structure
  2. Nov 11, 2026FactsFacts fetcher · Facts looked up
  3. Nov 11, 2026Entered the graphEntered the graph; copy stored