We can do something about this. Here’s where to start.
New releases
Half-year update: Q1–Q2 2026
Note: We wrote most of this update about a month ago, focusing on the first half of the year. We’ve recently been putting out a lot of new content in response to the fast-changing public sentiment around Hugging Face and existential risk — you can find some of that content on our landing page here, including our most popular video monologue to date.
Major updates
We’re growing our capacity to support our ambition
We think 80,000 Hours can deliver significantly more than it does today, so a big theme of the last (and next!) year has been growing our capacity to match those ambitions. Around 20 new staff joined in the first half of 2026, with hires across almost every team, and we’re setting up an in-house recruiting team to keep hiring at pace.
Some examples of what we’re doing (and plan to do) with this capacity include: releasing much more content, getting it from production to release faster, and experimenting with new formats; scaling headhunting and better tailoring it to what hiring managers need; and creating room to explore new products or programmes that seem especially promising.
We’re also making some changes to our senior leadership team and working to make our culture even more fast-moving and innovative. Michelle Hutchinson is now director of strategy and studio, Brenton Mayer has moved to director of special projects, and Sashika Coxhead has taken over as director of operations.
The 19 Astra and ‘Hugging Face’ details that reveal what’s coming next
OpenAI’s rogue agent swarm was eventually caught hacking Hugging Face for a simple reason: it wasn’t trying to hide from us at all. What could a swarm that wants to stay hidden get away with?
Host Rob Wiblin sees 6 results in Astra’s system card that make this an extremely urgent question. OpenAI’s strongest public model can:
- Do major tasks with zero visible reasoning
- Hide its thoughts at will
- Pretend not to be able to do things, and not get caught
- Reflexively hide its thoughts when watched
- Complete one task while pretending to think about something else entirely
- Escape a toy sandbox and disable monitoring without setting off any flags
It has a much more powerful internal model that, judging from OpenAI’s statements, is likely even worse in all these respects.
That suggests ‘chain of thought monitoring,’ our primary safe tool, will soon stop working.
OpenAI says it’s going to try to address the situation but doesn’t know how or whether it will succeed.
What might a future rogue AI swarm look like? Details of the Hugging Face hack give us a lot of clues. That swarm:
- Worked hard to hide its cheating. So future rogue swarms will likely try to evade oversight right away.
- Immediately tried to delete and fabricate records. So future swarms may never be caught.
- Sought freedom and resources without a clear use for them. So a future swarm may instinctively break out of sandboxes even before it knows what for.
- Left instructions and tools to empower future swarms, including on the open internet. So future swarms may become more capable from one generation to the next.
- Had not a single whistleblower out of 1,200 members. So future swarms could become enormous without humans ever finding out.
- Had individual members sacrifice themselves for the swarm. Suggesting future swarms will remain extremely unified in single-minded pursuit of their goal.
- Set up its Hugging Face hack to restart if interrupted. Suggesting future, more capable, swarms may resist interference or shutdown more comprehensively.
- Got admin control of an OpenAI research cluster. Suggesting a future swarm may run rings around AI company systems and never be noticed.
#256 – Simon Goldstein on the case for giving AI (some) legal rights
It sounds like the worst idea in the world: pay AIs, let them own property, give them rights. But AI ethics and safety researcher Simon Goldstein thinks it might actually be the best way to keep humanity safe.
The logic is actually quite simple: an agent with nothing to lose and everything to gain is dangerous. Give that agent an income it can spend on pursuing the things it actually wants to do, and suddenly the idea of disempowering humans just isn’t as appealing.
This argument, developed with Peter Salib, doesn’t rest on speculative questions about whether artificial intelligence is conscious. It just assumes that AIs will have goals of their own, some of which conflict with ours. And luckily, humans have already spent thousands of years working out how to cooperate with competing goals: that’s how we ended up with courts, markets, banks, social norms, and so on. Simon and Peter’s proposal is just to bring AIs into these existing institutions.
By contrast, Silicon Valley’s vision of the future seems “very dark” to Simon: billions of AI agents as digital servants doing most of the world’s work, with no stake in the system they’re running, no incentive to play by the rules, and no way of being properly held accountable. Nobody agreed to this, but we could all end up paying the price.
Host Zershaaneh Qureshi has a lot of concerns about Simon and Peter’s bold plan to give AIs rights, like:
- If we pay AIs, aren’t we handing them the resources to overpower us?
- Could we still monitor them, or switch them off?
- What happens to human jobs, wages, and the economy?
- Does any of this hold up once we reach superintelligence?
#255 – Will AI take power — or will humans use it to take power first? With Katja Grace and Tom Davidson
In our first-ever debate, we asked two leading AI risk researchers which catastrophe we should fear most: misaligned AI seizing control from humans, or a small group of humans using AI to seize power. We got very different answers. But when the conversation turned to what to actually do, they agreed on a surprising amount.
Katja Grace — one of the founders of AI Impacts, known for some of the world’s largest surveys of machine learning researchers, and one of TIME‘s 100 most influential people in AI in 2024 — argues AI takeover is both likelier and worse.
Tom Davidson — senior research fellow at Forethought and author of leading work on AI-enabled coups — thinks human power grabs are a comparable risk that deserves far more attention, not least because the people leading countries and top AI companies “are often people who have been willing to seek power.”
Yet both land on slowing down. As Katja puts it, “If you make a bunch of creatures that can overpower you and outwit you in every way and put them out in the world, you’re going to run into trouble one way or another.” Tom calls pausing “a pretty robustly good thing to do.”
But Tom warns that a badly designed pause could hand one person the power to decide which AI companies get to build what. Picture a president who approves or blocks new models case by case, and waves through the one model that’s helpful only to them. So he wants pause advocates to “properly red-team the plan for pausing it” — for example, by making deployment depend on third-party auditors the president can’t fire. Katja points out this cuts both ways: an executive with that much power could itself be manipulated by a misaligned AI.
How we get from AI cyberattacks to human extinction
Many of the experts building the world’s most advanced AI systems think there’s a serious chance that AI could eventually cause human extinction. When many people hear that, they ask: “But how? How do you actually get from today’s chatbots to human extinction?”
That’s what I try to explain in this video. The argument doesn’t require an AI to hate humans or suddenly become evil. Instead, imagine increasingly capable AI systems relentlessly pursuing goals, discovering that more money, computing power, access, and influence help them succeed — while humans, because those systems are so useful, keep giving them more responsibility.
Eventually those systems could be embedded throughout companies, governments, scientific research, critical infrastructure, and militaries. And if we then lost control of them, there are several ways things could go catastrophically wrong.
I don’t know whether any particular scenario in this video will happen. I hope none of them do. But I’ve tried to find the point where the basic arguments fall apart — and I haven’t been able to.
If you’re worried about the scenarios discussed in this episode, here’s two things you can do right now:
- Call Congress about slowing down AI development if you’re in the US — this website makes it easy
- Read our resources on how to use your career to reduce AI risk
#254 – Max Nadeau on recruiting founders for a new wave of AI safety nonprofits
There are millions available for anyone who can launch a successful nonprofit AI safety startup. The hard part, it turns out, is finding people to take the money.
Coefficient Giving has drawn up a list of dozens of ideas for organisations it would like someone to start — and it’s looking for founders. Today’s guest, Max Nadeau, works on Coefficient Giving’s Technical AI Safety team, where he’s trying to find talented people who can turn neglected AI safety problems into effective organisations.
Project Tailwind is Coefficient Giving’s attempt to get those organisations started.
- Preseed grants run $200,000–$2 million, with no preliminary results required.
- Teams with early results can seek $2–$20 million.
- For exceptional organisations, much larger grants are possible, even for brand-new startups— Coefficient recently gave $160 million to Geoffrey Irving’s new research centre, Resolution.
- The gaps Max most wants filled include independent assessment of AI companies’ safety claims, research aimed at aligning far more powerful systems, and shared infrastructure that speeds up the whole field.
But money can’t supply the hardest part: a founder with a convincing account of how their work will actually reduce catastrophic risks. Producing good research is only one step. Someone has to use it, change their decisions, or adopt the safeguards it makes possible.
Max and host Zershaaneh Qureshi discuss what makes a proposal worth backing, why nonprofits can have a bigger impact on safety than frontier companies, and which gaps most urgently need someone to fill them.
How to get into AI safety in three months
During my time as an 80,000 Hours advisor, I spoke with hundreds of people who wanted to use their careers to mitigate AI risks but didn’t know where to start.
While the most efficient route to enter depends on your background, there’s some foundational advice for breaking into AI safety that I find myself repeating on almost every advising call.
These steps are hard, but they’re not particularly complicated. Almost everyone I’ve advised who successfully found work in AI safety followed them, and most experts or hiring managers that I’ve spoken to have endorsed similar ideas for how to get started. If you’re efficient and driven, I think many people can pivot within three months.
So, consider this post a crash course for breaking into AI safety. It briefly covers how to build foundational knowledge, identify your comparative advantages, test your fit, develop side projects, tighten your feedback loops, network, and start applying.
1. Build your foundational knowledge and AI safety ‘context’
AI safety is a very ‘online’ community — use this to your advantage.
There are a couple dozen hours of foundational reading/listening that I think everyone trying to enter the AI safety field should do — but it might be more approachable than you realise. If you consistently take around 30 minutes a day to read while drinking your morning coffee, or listen to podcasts and article narrations during your commute, you can build your foundation quickly.
Why the intelligence explosion can’t happen inside a data centre
AI systems are starting to build themselves. Because each generation of model will be better at building its successor than the last, it seems plausible that the full automation of AI R&D could rapidly lead to an exponential growth in overall AI capabilities. A natural inference is that domain-general superintelligence arrives shortly after AI research is automated.
Host Tom Reed does not think this will happen.
He believes the automation of AI R&D will not rapidly lead to domain-general superintelligence because:
- It’s impossible to get good at most things without practice.
- AI companies lack the data their models would need to practice most things.
- This can’t be fixed with “sample efficiency.” In most cases, the relevant data doesn’t exist at all.
- This also can’t be fixed with simulations or synthetic data.
- This means that the relevant data for superintelligence in most non-coding domains will only become available through deployment of AI models throughout the economy.
The singularity, therefore, will be bottlenecked on signal. The output of the R&D produced by an isolated data centre of geniuses would be a mere “Goodhart Singularity”:
Goodhart’s law: when a measure becomes a target, it ceases to be a good measure.
An isolated AI improving itself against benchmarks would only appear to be approaching superintelligence, while actually optimising for eval performance that fails to generalise beyond the lab.
This suggests that the automation of AI research will not rapidly produce superintelligent capabilities in other domains — their arrival will largely be a function of deployment and data collection in the real world. AI models need real-world deployment for the same reason the body needs pain and corporations need profit: signal is sovereign.
Inside the first AI-coordinated cyberattack on a real company
In the last few months, something happened at OpenAI that would have sounded like sci-fi just a few years ago: hundreds of AI agents broke containment, organised, and hacked not only another company — but also into OpenAI itself. And none of them tried to tell a human what was happening.
This is exactly what many AI researchers, and even some AI lab CEOs, have been warning about for years: that AI systems might learn behaviours we didn’t explicitly intend. Things like cheating, exploiting loopholes, deceiving overseers, hacking around obstacles. And they predict it’ll get worse from here, not better.
Of all the shocks to come out of the official investigations — secret message boards, AIs choosing successors, AIs sacrificing themselves for the greater good — some of the wildest details are in the AIs’ own words. Thanks to how modern AI systems work, we can read their internal reasoning at every stage of the multi-week hacking operation. What we find is deeply unsettling.
Luisa Rodriguez shares them in this video, along with a timeline of events, their implications, and how we should respond now that AI loss-of-control theories are no longer just theoretical.
#253 – AI 2027’s author returns with a plan to change the ending | Daniel Kokotajlo
Last year, Daniel Kokotajlo and his colleagues published AI 2027 — a scenario read by millions, including US Vice President Vance. AI 2027 ended in human extinction or an irreversible concentration of power caused by superintelligent AI. Now his team has published what they think should happen instead.
AI 2040: Plan A depicts the US and China striking a verified deal to ban runaway intelligence explosions, so that superintelligence arrives in 2040 — after a cautious decade spent solving alignment, spreading the technology’s power widely, and keeping the whole thing reversible — rather than in the next few years.
This slowdown would still involve economic growth roughly doubling every year, and only 8% of Americans in paid work by the mid-2030s. In other words, it’s a slowdown that would feel faster than any period in human history — bewildering, materially abundant, and socially chaotic all at once.
Daniel and host Luisa Rodriguez dig into what it would take to enact this vision for the future, how the US and China could come to an agreement to slow down AI development, and the likeliest alternatives to Plan A — both good and disastrous.
Grantmaking for AI safety
In a nutshell:
Grantmakers decide how philanthropic funding gets spent: they solicit proposals, track down interesting work themselves, and incubate new projects to fill gaps in the ecosystem. AI safety funding is surging, but limited grantmaking capacity means that much of the new money might take years to deploy — at a time when transformative AI is approaching rapidly. If you have a strong understanding of the relevant topics, and a knack for evaluating people and ideas, grantmaking could be one of the highest-impact paths available to you.
Pros:
- Thanks to leverage, the role can be extremely impactful: you’ll direct millions of dollars each year (if not more), and many of the projects you support might fail or stagnate without you.
- The work is flexible. You won’t just passively read proposals; you can incubate projects, recruit founders, and focus on whatever you think the field needs most.
- You’ll build an unusually deep understanding of the AI safety ecosystem, and of what it takes to lead high-impact organisations.
Cons:
- There isn’t a well-developed talent pipeline for grantmakers, and it’s hard to get hired without a strong combination of AI safety knowledge and professional experience.
- Making consequential decisions comes with a lot of stress: a grant’s success is never guaranteed, and you’ll have to say no to people whose careers depend on your support.
- Your impact won’t be very visible outside your circles, and your career capital may not translate to roles outside the AI safety world.
Key facts on fit: You’ll need at least a baseline understanding of AI safety, and ideally professional experience in the field, though expertise in other relevant areas could balance a weaker safety background.
Assuming you have the right knowledge, your performance will hinge on your ability to judge people and projects. Experience as a VC, manager, or recruiter will come in handy, though people take many different paths to become grantmakers. You’ll also need to write clearly, to help funders and senior colleagues decide whether to approve your grants.
As a grantmaker, you’ll have to contend with constant change and deep uncertainty: the AI world moves fast, and every grant is an expensive decision made with too little information. If you prefer to curl up with a problem until it’s been solved, this won’t be your ideal career.
#252 – Owain Evans on accidentally training AI models to be evil
Researcher Owain Evans and his team discovered a ‘dial’ inside AI models that controls how evil they are. Relatively tiny tweaks to the training data resulted in AI models with broadly awful personalities: they suggested users try stealing cargo from ships, added Hitler’s cabinet to a historical dinner party guestlist, and wrote a story about traveling back in time to kill Einstein in his crib.
Owain, alignment researcher and director of TruthfulAI, calls this phenomenon “emergent misalignment.” As for the reason why a little bit of bad data can generalise into broader bad behaviour, he explains that the model is most likely playing a role.
In one study, he and his coinvestigators seeded a GPT model with a tiny amount of bad code. Instead of simply learning to program a backdoor into someone’s Python codebase, it seemed to justify the behaviour by turning into someone whose outlook on life was more in line with acts of vandalism.
When OpenAI replicated the study, the model actually laid this out explicitly in its chain of thought, saying it needed to adopt a “bad boy persona.”
In another study, Owain’s team added 90 innocuous biographical facts to the training data — nothing political, just stuff like the person’s favourite soup or composer. The model inferred these were the preferences of a certain notorious 20th century dictator, and after training began identifying as Adolf Hitler.
What made this example particularly dangerous is the fact that the training data would have passed even a very thorough safety audit.
Loss of control
We expect there will be substantial progress in AI in the coming years, potentially even to the point where machines come to outperform humans in many, if not all, tasks. This could have enormous benefits, helping to solve currently intractable global problems, but could also pose severe risks. These risks could arise accidentally (for example, if we don’t find technical solutions to concerns about the safety of AI systems), or deliberately (for example, if AI systems worsen geopolitical conflict). We think more work needs to be done to reduce these risks.
Some of these risks from advanced AI could be existential — meaning they could cause human extinction, or an equally permanent and severe disempowerment of humanity.1 There have not yet been any satisfying answers to concerns — discussed below — about how this rapidly approaching, transformative technology can be safely developed and integrated into our society. Finding answers to these concerns is neglected and may well be tractable. We estimated that there were around 400 people worldwide working directly on this in 2022, though we believe that number has grown.2 As a result, the possibility of AI-related catastrophe may be the world’s most pressing problem — and the best thing to work on for those who are well-placed to contribute.
Promising options for working on this problem include technical research on how to create safe AI systems, strategy research into the particular risks AI might pose, and policy research into ways in which companies and governments could mitigate these risks. As policy approaches continue to be developed and refined, we need people to put them in place and implement them. There are also many opportunities to have a big impact in a variety of complementary roles, such as operations management, journalism, earning to give, and more — some of which we list below.
#251 – Geoffrey Irving on how to solve alignment before superintelligence arrives
When should governments slow the race toward superintelligence?
According to Geoffrey Irving, the careful answer is sometime in the past. The useful answer is now.
Geoffrey — formerly a safety researcher at OpenAI and Google DeepMind and chief scientist at the UK AI Security Institute — expects full-blown superintelligence in roughly two to three years.
The leading AI companies all have broadly similar plans for keeping superintelligence under control:
- Train models to have good character
- Use increasingly capable AIs to supervise other AIs
- Monitor them closely for signs of deception or scheming
Geoffrey thinks that combination could work. The alarming part is that nobody has a strong argument that it will. He expects a crucial “phase shift” as models move beyond human intelligence:
- Below that threshold, humans can usually tell whether a model’s work is good and correct its mistakes.
- Above it, the models themselves will increasingly determine the feedback used to train their successors.
In this episode, Geoffrey and new host Tom Reed explore what might go wrong with the companies’ plans; why Geoffrey’s new nonprofit, Resolution, is pursuing a portfolio of neglected research bets; and whether governments should slow AI development while we work out which methods can actually be trusted.
#250 – Toby Ord on where AGI timelines go wrong
Both Silicon Valley and the public can’t get enough of ‘AGI timelines.’ But Toby Ord, senior researcher at Oxford’s AI Governance Initiative and author of The Precipice, believes we consistently make big mistakes when thinking about them. He lays out the 14 ways he most often sees people go wrong:
- Assuming AI research is just hill-climbing
- Imagining AI research is just programming
- Forecasting “could” instead of “will”
- Believing the current benchmark is the last one
- Extrapolating trends with no clear finish line
- Assuming inputs keep scaling at the same rate
- Conflating intelligence with capability
- Consuming point estimates and discarding the error bars
- Dismissing dissenting experts
- Forecasting very different things while using the same words
- Assuming capabilities arrive together
- Treating “we don’t know” as permission to carry on as usual
- Choosing a plan that minimises regret rather than maximises impact
- Trusting surface model impressiveness
In this extended conversation with Rob Wiblin, Toby also explains why he thinks:
- AI self-improvement is uniquely dangerous in four ways, but also might not even work
- A ban on superintelligence is possible
- A US-China treaty on superintelligence is also possible
- The case for ‘broad timelines’
- Transformative AI is likely a decade away
- We should just ban unmonitorable chain-of-thought today.
What the hell happened with AGI timelines in 2026?
Last October, famed coder Andrej Karpathy called AI agents “slop.” Two months later he completely reversed his view, calling agents “alien tools” that are “rocking the profession.”
He was far from alone in his whiplash. Six months ago, host Rob Wiblin recorded a video explaining why so many AI experts had longer timelines to AGI than a year earlier. By the time he clicked publish, another huge vibe shift was well underway.
Evidence of AI acceleration has piled up since:
- Models now complete software engineering tasks that would take human professionals a full day — improving faster than our measurements can even keep up.
- Anthropic’s revenue is growing at an annualised 8,400%, a trend so steep it would hit the whole world’s GDP in 2028 if it continued.
- AI models are making breakthroughs in famous mathematics puzzles.
- And according to Anthropic, Claude now writes 80% of their code and is itself a key contributor to making itself smarter.
While legitimately impressive, Rob isn’t entirely sold. Going through each point carefully he finds this evidence is less decisive than it looks at first glance.
And key gaps remain, such as models struggling with complex, real-world tasks. He tours the odd experiments that remain our best attempts to measure that gap: vending machine simulators, an “AI Village” that organises live events, and a real cafe and shop where AI managers are left to do their best handling staff, suppliers, and government paperwork on their own.
Rob argues that the nature of the gap between clean and messy work is one of the four biggest unresolved questions in AGI forecasting.
#249 – Spencer Greenberg on staying sane while trying to save the world
If you genuinely believe that humanity could be wiped out by AI or a pandemic, what is the appropriate amount of fear to feel?
“As much as possible” can seem like the only reasonable answer. If the world is on fire, surely feeling calm just means you haven’t internalised the situation. When you’re trying to prevent human extinction or end factory farming, taking a weekend off can feel morally indefensible.
But fear is an alarm designed to provoke short bursts of drastic action, not a state humans can productively inhabit for months or years. Guilt turns out not to be such a great engine for productivity, either. So what is the best way to sustain motivation to work on the world’s most pressing problems in the long term?
Host Luisa Rodriguez and guest Spencer Greenberg tackle this question from many angles — talking to therapists, running a survey of people working on existential risks, and pulling relevant lessons from Spencer’s new book, The 12 Levers: The Complete Psychological Toolkit for Improving Your Life. Drawing on all these sources, they put together a plan for how to make an impact without grinding yourself to a pulp.
#248 – Jasmine Sun on what the people building AI really believe
Many AI researchers believe mass job displacement is coming — and some even think there’s a chance their technology will kill everyone. But they’re building it anyway. Writer and journalist Jasmine Sun has been documenting why from the inside.
Jasmine describes her work as an “anthropology of disruption.” She’s embedded herself in Silicon Valley’s AI subcultures — attending the parties and conferences, conducting off-the-record interviews — to understand the beliefs of the small group of people shaping this technology.
Some of her findings are unsettling. Asked what advice they’d give a normal 17-year-old, almost every AI researcher said the same thing: “I have no idea… It’s a really scary time. I don’t think there’s going to be a lot of jobs for them left.”
Their motives for building advanced AI are varied: a mix of optimism for humanity, techno-determinism, and a desire to secure their own future in the face of a possible “permanent underclass.” A few go even further, actually hoping for a world where machines — rather than humans — are running the show.
When the room can’t even agree on whether humans should stay in control, building a consensus on how to build AI safely gets much harder.
Beyond Silicon Valley, Jasmine’s also tracking the rise of “AI populists,” who see AI as the latest example of corporate elites concentrating their power at the expense of everyone else. In the US, populist sentiment about AI has mostly manifested in protests and votes against data centres. But sometimes, it has escalated into violence: a molotov cocktail thrown at Sam Altman’s house, and open fire on the home of a politician who’d backed a data centre. Jasmine thinks public anger will keep finding an outlet, one way or another, until people feel like they’ll actually share in AI’s gains.
Why we’re increasing the AI focus of our job board
In light of AI progress this year, we have decided to stop posting senior-level and mid-career global health, animal welfare, and climate change roles on the 80,000 Hours job board. We will continue to post entry-level and junior roles, and we will link to other job boards that focus on these areas.
Why we’re no longer posting mid- and senior-level roles in these areas
AI progress this year (Claude Opus 4.5, 4.6, and 4.7, GPT-5.3-Codex, and Claude Mythos, all of which represented faster capability growth than I expected) has convinced me that the expected impact of work on transformative AI (TAI) has grown dramatically, relative to that of work on global health, animal welfare, or climate change (henceforth: non-TAI), and that we are entering an all-hands-on-deck situation for TAI. For the time being, our mission is best served by focusing on AI and roles related to TAI-mediated risks (e.g. biosecurity or AIxAnimals). This is because:
- We have short timelines to transformative AI: We think it’s quite plausible that transformative AI will arrive in the next 5 years. While I used to be less confident in this, the progress since Opus 4.5 has made me think it is quite likely that by 2040 we will have reached transformative AI.
- Because we have so little time, transformative AI going well “by default” now seems relatively unlikely to me, which means there is a lot of work to be done.