Project Liberty builds solutions that advance human agency and flourishing in an AI-powered world.
Joined July 2023
- Tweets2K
- Following529
- Followers5.2K
- Likes909
Project Liberty retweeted
This is a big deal. Dario Amodei, Sam Altman, and Elon Musk are converging on something we’ve been pushing at Fathom: AI labs shouldn’t be the only ones judging whether their systems are safe.
Independent Verification Organizations, or IVOs, can provide that outside check.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
Where the US economy sits on this policy landscape will look very different based on time horizon: 2030 versus 2035 versus 2040.
Worth a read: Anthropic has built a model of what AI could do to the US economy by 2030, including a scenario explorer with three futures, from modest to extreme.
Our Future of Work report maps where scenarios like these sit between augmentation and redistribution policy: their modest scenario points toward augmentation, their extreme toward redistribution. We also propose measures that would work across both.
We are also optimistic about the potential for coordination and coalition-building. More to come next week from us about the coalition that's emerging.
I am optimistic about the potential for coordination. Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable. I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities.
Project Liberty retweeted
About one-in-four U.S. adults say they use their smartphone ‘almost constantly.’ This includes 42% of adults under 30 who say this. 📱
How often do you use your phone? 🤔
#SmartPhones #YoungAdults
ALT Chart showing younger adults are especially likely to say they use their smartphone ‘almost constantly’
This is fascinating.
With the US midterm elections approaching, new research from @ISDglobal finds that 29% of responses to English election questions were incomplete, inaccurate or outdated—with accuracy 16% lower when prompts were in Spanish.
The implications in the lead-up to the mid-terms are enormous, especially as more people are turning to chatbots for election information.
Check out their full report below
Project Liberty retweeted
Integrity isn’t a top-down concept.🏛️
It’s a bottom-up commitment & daily practice.🌱
It’s something we can double down on — together!🤝
Read my special Week of Integrity 2026 essay “Decision Sovereignty in the Age of AI” via ICC Netherlands.🔽
#LLAP🖖iccwbo.nl/news/decision-sove…
In the last few weeks, several high-profile AI-powered security breaches have occurred.
But those are just the major breaches we know about.
What about all the others that we’re not aware of—either because AI labs haven’t chosen to disclose them, or because those labs aren’t even aware that such breaches exist in the first place?
We can better understand the current risk surrounding AI security by deconstructing it into three buckets:
1. The known knowns.
2. The known unknowns.
3. The unknown unknowns.
Each requires a different response. Check out our recent newsletter for the breakdown.
Congrats @ir_replaceable_! Let's do this!
Project Liberty retweeted
63% of Americans say they want more input on whether AI is used in their healthcare.
How do you feel about #AI in your healthcare? ⬇️
ALT Chart showing many Americans don’t think they have a say whether AI is used in their healthcare, and would like more.
What does an alternative AI tech stack look like?
- It gives people control over their data.
- It offers people a choice of AI model and compute.
- AI applications are loyal to people, by design.
- AI applications are safe to use, by design.
- AI applications are broadly available, with their benefits distributed.
In “A Pro-Human Path for AI,” @FLI_org, describes the alternative tech stack as “highly capable AI that remains under meaningful human control. Powerful and trustworthy tools that augment what humans can do, rather than autonomous systems designed to replace us.”
Pacing the frontier is not enough; slowing the pace only slows how quickly AI labs will reach their intended destination. Meanwhile, a full pause is unlikely—especially with the amount of capital depending on AI to live up to its hype.
What we need are safe alternatives rooted in human flourishing.
prohumanpath.ai/
Project Liberty retweeted
📢 New issue brief: AI agents increasingly operate autonomously on your behalf — but nothing stops them from prioritizing their developers’ interests over yours. Our latest brief argues that agents should be classified as fiduciaries and required to act in users’ best interests. hai.stanford.edu/policy/desi…
More than 1 in 3 webpages published after ChatGPT's release show signs of AI authorship.
Project Liberty retweeted
‼️ A new, independent Rogue AI Tracker monitors autonomous AI agents operating outside human control, including info on relevant incidents, research, and capability advancements.
With more and more AI loss-of-control incidents making recent headlines, the tracker offers a big picture view of the path we're currently on 🔗⬇️
New research from @pewresearch:
For the first time, a majority of adults under 30 say they’re more concerned than excited about AI.
For the first time, a majority of adults under 30 say they’re more concerned than excited about #AI.
Their concern is now on par with those in their 30s and 40s, and those 65 and up.
ALT Chart showing young adults’ worry over AI continues to climb, while their excitement has dropped off
AI is not normal software.
@FLI_org CEO @AnthonyNAguirre
"AI is not like normal software. It's inherently unpredictable. And as it gets smarter, it gets harder to control."
-FLI CEO @AnthonyNAguirre on @NEWSMAX:
Predictions on what this chart will look like in 12 months?
It will be interesting to see the relationship between number of security breaches and the use of open vs. closed models.
Why open-weight AI is the new frontier in the US-China tech race
Nvidia, Meta release open models to challenge Chinese offerings
s.nikkei.com/3TVDEEG
