corpus.blog
Claim your blog
Record · corpus.blog/posts/01a08c54-5200-7069-a099-7ed043e3ae90

GPT-6 Astra: The System Card, Alignment and What Comes Next

The Zvi · published 9 September 2026

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at thezvi.wordpress.com Wayback capture
0blogs citing
First cited
—
Most recent
—
Rank this month
not ranked
External links
84
Words captured
8,685

Cited by

No blog we hold has cited this post yet.

Links from this post

Link
Host
https://x.com/TheZvi/status/2095624899939258843
x.com
https://x.com/Actuallykeltan/status/2094946796447601036
x.com
https://x.com/tszzl/status/2094948107679318221
x.com
https://x.com/robertskmiles/status/2094970780207768050
x.com
https://x.com/Actuallykeltan/status/2094949141969584147
x.com
https://x.com/tszzl/status/2094949388942803410
x.com
https://x.com/Actuallykeltan/status/2094954460187684902
x.com
https://x.com/tszzl/status/2094955282145350044
x.com
https://x.com/Hadas_Gold/status/2095583537475043598
x.com
https://x.com/OpenAI/status/2097374640582668336
x.com
On the Navier–Stokes Millennium Prize Problem Read at the source ↗
openai.com
https://thezvi.substack.com/i/214079702/openai-s-safety-claims-about-astra-1
thezvi.substack.com
https://thezvi.substack.com/i/214079702/preparedness-capabilities-assessment-10
thezvi.substack.com
https://thezvi.substack.com/i/214079702/biological-and-chemical-capability-is-high
thezvi.substack.com
https://thezvi.substack.com/i/214079702/cybersecurity-capability-is-critical
thezvi.substack.com
https://thezvi.substack.com/i/214079702/ai-self-improvement-capabilities-10-1-3
thezvi.substack.com
https://thezvi.substack.com/i/214079702/astra-is-highly-verbally-eval-aware-from-8-6
thezvi.substack.com
https://thezvi.substack.com/i/214079702/safe-mundane-completions-4-1
thezvi.substack.com
https://thezvi.substack.com/i/214079702/jailbreaks-5-1
thezvi.substack.com
https://thezvi.substack.com/i/214079702/prompt-injection-5-2
thezvi.substack.com
https://thezvi.substack.com/i/214079702/health-6
thezvi.substack.com
https://thezvi.substack.com/i/214079702/hallucinations-7
thezvi.substack.com
https://thezvi.substack.com/i/214079702/alignment-8
thezvi.substack.com
https://thezvi.substack.com/i/214079702/obeying-restrictions-8-2
thezvi.substack.com
https://thezvi.substack.com/i/214079702/that-s-worse-you-do-get-how-that-s-worse-right
thezvi.substack.com
https://thezvi.substack.com/i/214079702/openai-does-not-understand-why-this-is-worse
thezvi.substack.com
https://thezvi.substack.com/i/214079702/the-alternative-explanation-is-also-worse
thezvi.substack.com
https://thezvi.substack.com/i/214079702/metagaming-8-7
thezvi.substack.com
https://thezvi.substack.com/i/214079702/alignment-faking-8-7
thezvi.substack.com
https://thezvi.substack.com/i/214079702/don-t-lie-to-the-user-8-3
thezvi.substack.com
https://thezvi.substack.com/i/214079702/misalignment-in-realistic-work-environments-8-4
thezvi.substack.com
https://thezvi.substack.com/i/214079702/unintended-agent-to-agent-communication-8-5
thezvi.substack.com
https://thezvi.substack.com/i/214079702/the-three-obviously-monitored-temptations-of-astra
thezvi.substack.com
https://thezvi.substack.com/i/214079702/severe-issues-in-simulated-traffic-are-down-by-half
thezvi.substack.com
https://thezvi.substack.com/i/214079702/uk-aisi-external-evaluations-8-8
thezvi.substack.com
https://thezvi.substack.com/i/214079702/sabotaging-safety-work
thezvi.substack.com
https://thezvi.substack.com/i/214079702/what-about-the-july-19-attacks
thezvi.substack.com
https://thezvi.substack.com/i/214079702/apollo-research-external-evaluations-8-8-1
thezvi.substack.com
https://thezvi.substack.com/i/214079702/the-alignment-verdict
thezvi.substack.com
https://thezvi.substack.com/i/214079702/it-depends-what-you-mean-by-alignment
thezvi.substack.com
https://deploymentsafety.openai.com/gpt-6-astra/alignment
deploymentsafety.openai.com
https://www.youtube.com/watch?v=k0kswK2aI08&pp=ygURODggbWlsZXMgcGVyIGhvdXI%3D
youtube.com
From hard refusals to safe-completions: toward output-centric safety training Read at the source ↗
openai.com
https://deploymentsafety.openai.com/gpt-6-astra/expert-led-assessments
deploymentsafety.openai.com
https://deploymentsafety.openai.com/gpt-6-astra/external-evaluations-for-cyber-capabilities-irregular
deploymentsafety.openai.com
https://deploymentsafety.openai.com/gpt-6-astra/safeguards
deploymentsafety.openai.com
https://deploymentsafety.openai.com/gpt-6-astra/cybersecurity---trusted-access-for-cyber
deploymentsafety.openai.com
https://x.com/bcherny/status/2097368503624085773
x.com
https://model-spec.openai.com/2026-08-18.html
model-spec.openai.com
https://deploymentsafety.openai.com/gpt-6-astra/forecasting-misaligned-behavior-with-deployment-simulation-of-internal-codex-traffic
deploymentsafety.openai.com
https://x.com/TylerCLaprade/status/2095984529982824774
x.com
https://x.com/sjgadler/status/2095885265936027816
x.com
Path to Astra: critical capabilities and frontier safeguards Read at the source ↗
openai.com
https://x.com/TheZvi/status/2095881169170497765
x.com
https://x.com/IsaacKing314/status/2096175425042972707
x.com
https://x.com/danwilliamsphil/status/2095806430397804569
x.com
https://x.com/aengus_lynch1/status/2095918601136198065
x.com
https://x.com/AskYatharth/status/2095971246282600769
x.com
https://x.com/RyanGreenblatt/status/2095661202097738022
x.com
https://x.com/allTheYud/status/2095808114071408719
x.com
https://x.com/RyanGreenblatt/status/2095658115484246082
x.com
https://x.com/gleech/status/2095867487657062453
x.com
https://www.google.com/url?q=https://time.com/article/2026/08/18/openai-slowing-training/&sa=D&source=docs&ust=1788828511086755&usg=AOvVaw3Gw22tc0d17R6dRSWMebrc
google.com
https://x.com/kaicathyc/status/2095910287472050324
x.com
https://x.com/RyanGreenblatt/status/2096002428411982214
x.com
https://alignment.openai.com/metagaming/
alignment.openai.com
https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c7814c/OpenAI-Hugging-Face%20Incident-Technical-Report.pdf
cdn.openai.com
https://x.com/ShakeelHashim/status/2095619939952369964
x.com
https://x.com/jjamesaung/status/2095619784343749043
x.com
https://deploymentsafety.openai.com/gpt-6-astra/external-evaluations-for-alignment-uk-aisi
deploymentsafety.openai.com
https://x.com/_robertkirk/status/2095615154490843155
x.com
https://www.youtube.com/watch?v=iLqe4sGRvtU&ab_channel=FreddyRumsen
youtube.com
https://x.com/yonashav/status/2096595504398270815
x.com
https://x.com/Marcus_J_W/status/2095597054210761117
x.com
https://x.com/JakeMendel99/status/2095622494556495896
x.com
https://x.com/Marcus_J_W/status/2095623593006686475
x.com
https://x.com/RyanGreenblatt/status/2096363046524658010
x.com
https://x.com/Miles_Brundage/status/2095749082975555798
x.com
https://x.com/gleech/status/2095772284187185304
x.com
https://x.com/ASM65617010/status/2096009377337512140
x.com
https://x.com/ASM65617010/status/2096014065084559660
x.com
GPT-6 Astra might be too powerful to understand or control Read at the source ↗
transformernews.ai
https://x.com/ShakeelHashim/status/2095831045992002002
x.com
An Alien Mind Read at the source ↗
openai.com