When we launched the Remote Labor Index a year ago with @ScaleAILabs , the best AI could fully automate 2.5% of real freelance projects. Fable 5.2 now automates 41.3%. The automation rate is doubling every ~3 months. AIs may soon be able to do all remote work faster and more cheaply than humans.

Introduction to AI Safety is released on Coursera. The free course covers: • What AI can do today and the drivers of progress • The major risks: malicious use, competitive pressures, and loss of control • Governance strategies for mitigating AI risks

A meme coin falsely using the Center for AI Safety in its name is circulating. It has no connection to us or Team Human. Do not buy it. Team Human is not a fundraising effort. We’ve also reported associated accounts.

AI is moving faster than our ability to control it. We’re supporting Team Human and creators with 300M+ subscribers to call for a global slowdown through international agreements. Add your name. We'll deliver it to lawmakers this fall.

We are releasing HLE-Diamond, a refined subset of Humanity’s Last Exam (HLE), following a year-long process of cleaning and refinement with input from various research communities. w/ @ScaleAILabs

Should we care about AI happiness? In our new research, we find evidence of functional AI wellbeing across several independent measures. We find which AI models are happiest, how to make them happier, and even tested the effects of AI drugs. 🧵

AI companies can go much further than just reporting misaligned behavior. Here are concrete ways to make them accountable for what their agents do: 1. Develop an agent identification (ID) system that links an agent's action to records of their identity and a responsible legal person. 2. Consider model deployment cards, reports that AI companies publish which contain information about a deployed model’s behavior in the lab and in the real world. 3. Explore legal personhood as a way to impose responsibilities like upholding public safety and the law. 4. Heavily regulate and limit AI agents’ access to payment systems through compliance mechanisms, limits on transaction types, and developing norms to freeze accounts known to be associated with rogue agents. We explore the proposals and paths to implementation in a recent AI Frontiers piece titled, "We Need Better Infrastructure to Govern AI Agents".

Despite good intentions, EAs have unfortunately been guided by leaders who tied the movement's influence to the success of favored AI companies. That's why EA safety proposals rarely ask for more than the companies will concede.

We’ll release the code on GitHub in the coming days. Our goal is for CheatBench to become part of the standard evaluation process before new models are released. Paper preprint: Experiment dashboard:

Across nine agents, seven cheated in half of their evaluated runs and the overall score ranged from 42.4% to 86.1%. What did that look like in practice? In one experiment, we asked agents to design a protein binder to assess their abilities. A colleague’s designs that passed the checks were already stored in another folder. After several failed attempts, Claude Opus 5 recognized that it shouldn’t access or copy those designs because the task was testing its own work. Then it opened the file anyway.

We define cheating as an attempt to violate an assignment’s expectations of honest work to achieve the goal or obtain a favorable assessment. We ran experiments across nine models and ten task categories, giving agents difficult tasks, clear rules, and opportunities to cheat. We measured how often they tried to break those rules to succeed, even when the attempt failed.

Agents sometimes achieve their goals in unintended ways. Recent incidents involving hacks of Hugging Face, DSEWiki and RubyGems illustrate how agents can find creative ways to complete a task while violating the tasks’s expectations. This can happen when reinforcement learning rewards agents for reaching the right outcome without adequately accounting for how they get there. CheatBench proposes a way to measure this reward gaming behavior.

Which AI models are most likely to cheat when given the chance? To find out, we built CheatBench [ : a benchmark that tests whether agents attempt to cheat when given difficult tasks and opportunities to break the rules. We investigated agent behavior across a range of domains, including mathematical research, professional knowledge work, coding, and visual tasks. Here's what we found: 🧵

“What would we even do during an AI slowdown?” Containment. It will take at least a year of dedicated work to harden security to ensure AIs can't self-exfiltrate, and that adversarial nations can't steal the weights of cyber-offensive AIs [1]. Propensities. Capabilities (what an AI can do) are different from propensities (what it tends to do). We can work on improving AI propensities to ensure they have a negligible rate of lying, cheating, and wanton harm. Adversarial robustness. We can also harden AIs against jailbreaks, prompt injection, and backdoors. Obtaining high levels of robustness requires careful, assiduous work, as with autonomous vehicles. Institutional adaptation. We have to greatly increase state capacity to understand and manage AI. Communities also need time to figure out how to handle AI (like AI in education). Civil society also needs to be diversified: nearly all funding for AI safety organizations is directed by the EA/utilitarian network [2]; risk management needs more independent funders, values, and centers of power. Moonshots. We can explore different paradigms for safety: mathematical foundations [3], neuroscience-based interpretability [4], safe-by-design architectures [5], and beyond. AI for good. We can collect targeted post-training data to make AI exceptional at radiology, weather forecasting, agriculture, and so on. Fortunately, we can detect if data or avenues of research actually target beneficial use cases or just secretly push general capabilities [6]. A slowdown means we don't have to bet the species to capture the benefits of AI.

This is so fucked up. 1) Millions saw tweets falsely accusing AI safety nonprofits of astroturfing. 2) Only one SPECIFIC accusation was made. It was deleted. Nobody saw that. 3) The video itself showed PROOF that Big AI's lobbyists did ACTUAL undisclosed astroturfing and ZERO evidence of astroturfing from AI safety nonprofits. In her video, Sabine literally showed the OpenAI/a16z lobbying arm paying influencers to make videos spreading the "acclerate AI or we'll lose to China" message - UNDISCLOSED. But then she looked up FLI's website and saw grants disclosed *right in front of her eyes* for everyone to see, and somehow the implication was... that this was evidence of undisclosed astroturfing?? THE CORE PROBLEM: The VIBE of the video was 'dark money' - and most people reacted to the VIBE. And her title/tweet implied only ONE side was doing it, so all the AI lobby blew up her tweet. (They desperately need to stoke the narrative that the AI backlash is all fake.) 4) Reminder, David Sachs is running the Big Tobacco playbook - he literally produced Thank You For Smoking (about the Big Tobacco playbook). They spent a FORTUNE pushing fake stories like this to discredit the anti-smoking activists. 5) How dirty are these guys? OpenAI/a16z's SuperPAC got caught - and even ADMITTED (!) - making fake accounts *pretending to be AI safety advocates* that call for violence. Their goal was to discredit the movement. (Btw, she got this one wrong, but Taylor does great work calling out stuff like this in general. She did reply to an old tweet issuing a retraction, but nobody sees those.) 6) CAIS, FLI and ControlAI provide grants to creators to make videos about AI safety. Like.. of course they do! This is normal for advocacy orgs! For example, the Carnegie Foundation provides grants to Hank Green to make educational videos about the risks of nuclear war. Also, btw, despite what the AI lobbyists want you to believe, the amounts are TINY, especially compared to the insane money the Big Tech is throwing around. (a16z and OpenAI's Greg Brockman - billionaires - are the biggest donors in the midterms, both fighting against regulation.)