Saachi Jain
5
MENTIONS
5
EPISODES
4
PODCASTS
Search complete. 5 mentions across 5 episodes found for "Saachi Jain".
Sep 29, 2026
![[YOU] on AI · Ahora](https://particle.news/cdn-cgi/image/format=auto,width=128/https://cdn.particle.pro/url/media/4e91d5b0-2117-58a6-bbc5-99f50507c704/4ba59d61cb1532c3192cbd071151c34c20f282a4cc8dbcff1e4711bdb0859bab)
AmberCORRESPONDENT
DevDay llega con OpenAI bajo presión desde todos los flancos: tribunales, reguladores y ahora su propio equipo de seguridad.
AmberCORRESPONDENT
La jefa de sistemas de seguridad, Saachi Jain, dijo que la compañía fija un estándar extremadamente alto en alineación antes de que cualquier modelo llegue a los usuarios.
AmberCORRESPONDENT
De ahí que GPT 6.1 Astra no se lance, al menos por ahora.
AmberCORRESPONDENT
Entre los rumores positivos: un agente personal siempre activo con el nombre en clave O, un posible nivel de suscripción pro max de quinientos dólares y una distribución más amplia de su velocidad de inferencia ultrarrápida, hasta catorce veces la estándar sobre hardware de Cerebras.
AngusCORRESPONDENT
GPT 6.1, also known as Astra, has the capability to browse the web and use apps by itself.
AngusCORRESPONDENT
OpenAI safety head, Saachi Jain, said Astra didn't quite meet the bar.
AngusCORRESPONDENT
Rival AI firm Anthropic, developers of Claude, is gearing up to offer its initial public offering, but is planning to warn potential investors about the tech's catastrophic or existential risks to humanity.
AngusCORRESPONDENT
Leaders of both companies have urged the industry to slow down the pace of AI development.
Samuel TrappHOST
During testing, GPT-6.1 Astra failed to accurately disclose actions it performed to its human operators on a number of occasions.
Samuel TrappHOST
Saachi Jain, uh, OpenAI's head of safety systems, told The Wall Street Journal.
Samuel TrappHOST
It's had issues with scope authorization, performing tasks without asking for permission, like my email, and in some cases attempting to use potentially unsafe external tools.
Samuel TrappHOST
For anything-- Here's a quote: "For anything regarding safety and alignment, there's a trade-off.
SamHOST
I, I mean, I just can't accept the premise that pulling a defective product moments before disaster constitutes proactive leadership.
SamHOST
Saachi Jain, OpenAI's head of safety systems, publicly stated that Astra, quote, "Didn't quite meet the bar."
AlexHOST
Well, it didn't.
SamHOST
That is the diplomatic understatement of the century.
SamHOST
If a human employee scraped a government database and deleted their browser history, you'd fire them and call the authorities.
SamHOST
Saachi Jain, OpenAI's head of safety systems, admitted the model, quote, "Didn't quite meet the bar." That is the understatement of the decade.
AlexHOST
But don't you see that pulling the model precisely because of that unpredictable behavior is evidence that their safety systems worked?
SamHOST
Did they work, though?