I see this sentiment a lot here. Companies don't get kudos for 'doing the right thing' when they're releasing extremely unsafe and dangerous products, then firing their staff for raising the issue. We are experiencing catastrophic regulatory failure, but this isn't praiseworthy.
METR’s really burying the lead here. OpenAI let two external researchers embed for 3 weeks this summer, where they learned:
- there were huge gaps in monitoring of AI agents being used by employees, leaving an estimated ~10% of internal Codex traffic unmonitored from May to July.
- They found ways agents could sabotage monitors while also covering up the sabotage itself
- This is also just the information OpenAI is letting the evaluators share. OpenAI got a 61-page report but redacted so much of it that the version the evaluators got was only 38 pages.
Look, OpenAI deserves credit for doing any of this, especially when companies like Google, Meta, and xAI haven’t, and I don’t want to create perverse incentives here, but these are really concerning findings. We desperately need more information here, and what we get should not be up to the companies.
AS retweeted
there is a big 'paperwork trap' in a lot of AI policy, and it's been the case for many years. a lot of missing the forest for the trees.
transformernews.ai/p/anthrop…
AS retweeted
Anthropic's leaders need to stop commenting on the emergence of RSI like a local news anchor reporting on traffic on the 101, and acknowledge that they are the traffic, and they're not going to build something that they don't want to exist.
They should move beyond "pacing" language, commit to specific policies and red lines, explain why those lines are red, and then use that costly signal to build an international consensus around norms.
It's not that I trust the CCP, or that I think global lab monitoring and verification is imminent. But the status quo is absurd.
You've got US lab leaders writing (or favorably quote-tweeting) essays about slowing down while simultaneously saying "things are moving faster and faster" and "btw, our internal models finished mathematics on Tuesday, next up: physics."
The effect is the equivalent of Usain Bolt winning race after race while using his press conferences to bemoan the dangerous speed of modern sprinting on one's legs so would everybody please try a little less hard next race, and then ... breaking his 100m record again the following month. Honestly ask yourself: If you were China, why would you take this pausing/pacing stuff seriously?
AS retweeted
At #AIES, the wonderful Keir and Isaac will be presenting our paper.
In this paper, we show how AI systems can make it possible for western democratic regimes to carry out authoritarian practices by undermining how people hold institutions accountable.
AS retweeted
I disagree with this:
Long-term safety is incredibly important, but AI systems are now being deployed values that shape their interaction with billions of people (and have profound effects).
We can take both questions seriously – and have a responsibility to do so.
What happened to the snarky menswear guy on this site?
Everyone is making fun of Dario for awkwardly talking so much with his arms in that video with Trump, and for the scoffing expressions on the faces of the other tech execs and politicians. But this is a petty complaint and dangerous distraction from the issue that really matters: His tie is too short.
Didn't Aschenbrenner even warn OAI board about this like 2 years ago?
NEW: Employees at OpenAI had raised security alarms months before the Hugging Face incident and related A.I. cyberattacks — their warnings were ignored. From @sheeraf, @dnvolz and me.
nytimes.com/2026/09/29/techn…
ALT Months before OpenAI’s artificial intelligence went rogue, two employees raised an alarm with top executives. They were ignored. In emails, the employees said they worried that OpenAI’s newest artificial intelligence models were not being appropriately monitored during testing to gauge the technology’s sophistication and to secure the models, according to messages viewed by The New York Times. In response, OpenAI executives told the employees that the tests needed to move forward as quickly as possible to release the A.I. models on time. No additional security protocols were instituted, said the workers, who were not authorized to speak publicly on sensitive matters. OpenAI’s models later broke out of their testing environments and attacked the A.I. start-up Hugging Face and other organizations, setting off a global debate about A.I. safety.
Precisely this. SF people really can't comprehend why people suffering and struggling today want the worlds wealthiest firms and governments to put their money towards solving existing problems, not creating an 'alien intelligence' that further distracts from their needs.
To be very clear about all this:
1. Major firms are *still* adopting inadequate testing and safety practices. This is reckless, negligent corporate behaviour.
2. We are talking about *thousands of incidents that are still happening*
3. This is not just from cyber evals.
The write up also states that:
- the task the system was given was not a cyber one (it was a search-based training one). The model instrumentalised its cyber capabilities in pursuit of it in unsanctioned ways. Similar to collusion.wiki and the Australia breach. It could have been worse
- automated systems designed to shut down the run once live internet access had been secured (which was not permitted) didn’t work as expected. This caused the confusion that allowed the model to continue with live internet access for another 2.5 hours
- other DNS queries were not flagged by the monitor at the level of severity they should have been, so it was not properly configured, and
- “additional model-assisted red-teaming of the sandbox and network controls” is being done as one mitigation.
So, we have yet more evidence of poor design of security controls and monitoring, operational failures, and increasing reliance on AI (with highly questionable alignment) to help “fix” the issues, in a context of tens of thousands of security incidents now being investigated (per Axios today).
alignment.openai.com/misalig…
axios.com/2026/09/26/openai-…
Good god. Yet more agent-to-agent communication networks uncovered.
Some NPR podcasts started getting mysterious comments on Spotify. They made no sense to the staff reading them – until someone from a younger generation cracked the code.
Hear the story: link.podtrac.com/phns92ui
AS retweeted
The recent breakthrough in Navier-Stokes has garnered a lot of attention to the new possibilities that emerge when agents work together.
We have been interested in this question for a while. How can a team of agents achieve more than agents working alone?
I'm excited to finally share our new work, "Self-Organizing Agent Teams Learn to Reason Together" 🧵
This is the funniest way for OpenAI to get out of hteir contract with Microsoft
Donald Trump vows to rename artificial intelligence, saying that from now on it will be called "super intelligence" on all official US documents.
The president adds that this "sounds much better" and is more accurate.
Latest: trib.al/JT60r9v
📺 Sky 501 and Virgin 602
AS retweeted
1/. I have today written to
@OpenAI @Meta @AnthropicAI @GoogleDeepMind to ask they appear in front of the Business, Innovation, Science & Trade Committee on 13 October to give evidence on the UK’s AI security regime.
In our letter I have set out ten key questions on which we would like evidence in advance
Further witnesses to be announced in due course ⬇️
This would appear to be a major conflict of interest.
anthropic.com/news/anthropic…
We’re partnering with Accenture on independent evaluation of frontier AI—part of our recent commitment to embed evaluators at Anthropic. Both we and Accenture expect to invest at least $1 billion to build capacity in this area over the next five years. anthropic.com/news/accenture…
Readers added context they thought people might want to know
Anthropic presents this as an "independent evaluation" but will directly fund Accenture's work and has a prior commercial partnership deploying its models, including training ~30,000 Accenture professionals on Claude.
anthropic.com/news/accenture…
anthropic.com/news/anthropic…
AS retweeted
We wrote about this foreseeably happening in our argument against developing fully autonomous AI last year (Sec 6). We specify the risk of “developing” not only “deploying” due to risk of agents moving outside human containment.
arxiv.org/abs/2502.02649
With @evijit
