🧵Today we’re sharing more details about improvements of the default GPT-5 model in responding to sensitive conversations around potential mental health emergencies and emotional reliance. These changes reflect the careful work of many teams within OpenAI and close consultation with experts - including more than 170 mental health clinicians.

This is interesting. Gpt-5-chat-latest quietly shot to the top of spiral bench. This is the model served on chatgpt dot com, though I test it via api so we avoid any safety routing. I'm inferring this is due to the oct 3 update since they don't version this model.🤔

Earlier this month, we updated GPT-5 with the help of 170+ mental health experts to improve how ChatGPT responds in sensitive moments—reducing the cases where it falls short by 65-80%. openai.com/index/strengtheni…

Valthos builds next-generation biodefense. Of all AI applications, biotechnology has the highest upside and most catastrophic downside. Heroes at the frontlines of biodefense are working every day to protect the world against the worst case. But the pace of biotech is against them: more powerful methods to design biological systems, with near-universal access, open up an increasing surface area of threats. In this new world, the only way forward is to be faster. So we set out to build the tech stack for biodefense. Our team of computational biologists and software engineers applies frontier AI to identify biological threats and update medical countermeasures in real-time. We are backed by $30M from @OpenAI, @Lux_Capital, @foundersfund and others including @Definition_Cap. We are actively hiring engineers to join in the mission - if that sounds like you, get in touch.

There’s quite a lot more to the story than this. As everyone knows, we are actively defending against Elon in a lawsuit where he is trying to damage OpenAI for his own financial benefit. Encode, the organization for which @_NathanCalvin serves as the General Counsel, was one of the first third parties - whose funding has not been fully disclosed - that quickly filed in support of Musk. For a safety policy organization to side with Elon (?), that raises legitimate questions about what is going on. We wanted to know, and still are curious to know, whether Encode is working in collaboration with third parties who have a commercial competitive interest adverse to OpenAI. The stated narrative makes this sound like something it wasn’t. 1/ Subpoenas are to be expected, and it would be surprising if Encode did not get counsel on this from their lawyers. When a third party inserts themselves into active litigation, they are subject to standard legal processes. We issued a subpoena to ensure transparency around their involvement and funding. This is a routine step in litigation, not a separate legal action against Nathan or Encode. 2/ Subpoenas are part of how both sides seek information and gather facts for transparency; they don’t assign fault or carry penalties. Our goal was to understand the full context of why Encode chose to join Elon’s legal challenge. 3/ We’ve also been asking for some time who is funding their efforts connected to both this lawsuit and SB53, since they’ve publicly linked themselves to those initiatives. If they don’t have relevant information, they can simply respond that way. 4/ This is not about opposition to regulation or SB53. We did not oppose SB53; we provided comments for harmonization with other standards. We were also one of the first to sign the EU AIA COP, and still one of a few labs who test with the CAISI and UK AISI. We’ve also been clear with our own staff that they are free to express their takes on regulation, even if they disagree with the company, like during the 1047 debate (see thread below). 5/ We checked with our outside law firm about the deputy visit. The law firm used their standard vendor for service, and it’s quite common for deputies to also work as part-time process servers. We’ve been informed that they called Calvin ahead of time to arrange a time for him to accept service, so it should not have been a surprise. 6/ Our counsel interacted with Nathan’s counsel and by all accounts the exchanges were civil and professional on both sides. Nathan’s counsel denied they had materials in some cases and refused to respond in other cases. Discovery is now closed, and that’s that. For transparency, below is the excerpt from the subpoena that lists all of the requests for production. People can judge for themselves what this was really focused on. Most of our questions still haven’t been answered.

One Tuesday night, as my wife and I sat down for dinner, a sheriff’s deputy knocked on the door to serve me a subpoena from OpenAI. I held back on talking about it because I didn't want to distract from SB 53, but Newsom just signed the bill so... here's what happened: 🧵

We’re updating GPT-5 Instant to better recognize and support people in moments of distress. Sensitive parts of conversations will now route to GPT-5 Instant to quickly provide even more helpful responses. ChatGPT will continue to tell users what model is active when asked. Starting to roll out to ChatGPT users today.

Excited to share details on two of our longest running and most effective safeguard collaborations, one with Anthropic and one with OpenAI. We've identified—and they've patched—a large number of vulnerabilities and together strengthened their safeguards. 🧵 1/6

No single person or institution should define ideal AI behavior for everyone.  Today, we’re sharing early results from collective alignment, a research effort where we asked the public about how models should behave by default.  Blog here: openai.com/index/collective-…