@Miles_Brundage: This was my reaction to that OAI security post also. No one (informed) denies there are good security people there, the…
Summary
A tweet discusses concerns about OpenAI's culture and decision-making processes regarding security, highlighting externalization of costs.
View Cached Full Text
Cached at: 09/29/26, 03:46 AM
This was my reaction to that OAI security post also.
No one (informed) denies there are good security people there, the concern that informed people have about OAI and other companies right now is about culture and decision-making processes
Sean Heelan (@seanhn): Or, to put it another way, obviously it’s not just the “fucking sandbox”. It’s a culture/leadership issue.
There is some non-zero number of incidents that OAI execs are willing to tolerate. And every one of those is OAI externalising costs instead of internalising them.
Similar Articles
@Miles_Brundage: Especially galling when OAI is under fire for recent incidents that are obviously more about sloppiness than any kind o…
Miles Brundage criticizes the backlash toward OpenAI, arguing that recent incidents stem from operational sloppiness rather than any fundamental difference in philosophy.
@Miles_Brundage: Not familiar w/ the details here but will repeat my refrain on this kind of thing: This being news that more than a few…
Miles Brundage comments that the news about OpenAI's safety head leaving being widely concerning indicates a policy failure, and calls for industry-wide safety standards and audits.
@Miles_Brundage: This is why I am so critical of Anthropic’s anti-OAI identity and OAI’s recent “Anthropic is authoritarian” BS Neither …
Miles Brundage criticizes the public antagonism between Anthropic and OpenAI, arguing that their anti-each-other stances make meaningful coordination harder and risk justifying corner-cutting in the AI race.
@Miles_Brundage: *taps the sign* https://milesbrundage.substack.com/p/why-security-comes-first… Also RAND had a great report on protecti…
Miles Brundage argues that security is a critical priority for frontier AI companies to prevent theft and tampering, highlighting current vulnerabilities and the need for immediate action.
@Miles_Brundage: Surprised to not see more commentary on this excerpt from the Opus 5.5 blog post: "We largely understand the risks toda…
A tweet critiques a statement from Anthropic's Opus 5.5 blog post, arguing that the claim about understanding and managing today's AI risks is obviously false.