Your next steps

Choose your focus

Create your 1-pager

Next steps: Technical fellowships

Next steps: Policy fellowships

Next steps: Other fellowships

Thank you!

How does AI think?

Interpretability in practice

Making AI go well

What might success look like?

Building AI safely is hard

What future do you want?

Assuming harm

Building defences

Break the kill chain

Evaluations: AI can, but will it?

How do AI companies test for safety?

Option 1: Anthropic

Option 2: OpenAI

Option 3: Google DeepMind

Option 4: Meta

Can we train AI to be safe?

Feeding AI ‘good’ data

Teaching AI right from wrong

More safety techniques

<Embed url="https://www.youtube.com/embed/MCixBdLkmpQ"/>

**You're at the beginning of your technical AI safety journey.**

This course gave you a structured overview of technical AI safety: the key challenges behind making AI safer, the current safety techniques, and their gaps. 

You’ve started developing some **research taste** — the ability to assess which areas of technical AI safety are impactful to work on and why. With &lt;2,000 people working full-time on making AI systems safer (based on our estimates), the field needs people with **[high agency](https://usefulfictions.substack.com/p/how-to-be-more-agentic)** – those willing to take bold action even when it’s hard.

There’s no clear roadmap, instructions or prophets to tell us what we should do – just people who care deeply, doing what they believe will steer us toward a positive future. You don't need a perfect [inside view](https://www.neelnanda.io/blog/47-inside-views) on AI safety to start. This will develop as you experiment. In the meantime, lean on open problems and proposals from organisations you respect.

Many have written great advice on technical AI safety careers, like[ 80,000 Hours](https://80000hours.org/career-reviews/ai-safety-researcher/), [Marius Hobbhahn](https://www.alignmentforum.org/posts/kpmaEevZ2KehZo2tp/some-advice-on-independent-research) (CEO and Founder of Apollo), [Neel Nanda](https://www.alignmentforum.org/posts/jP9KDyMkchuv6tHwm/how-to-become-a-mechanistic-interpretability-researcher) (leading mech interp researcher at Google DeepMind). This synthesises that advice into specific actions you can take after this course.

The most common advice we give to people is:

**1: When in doubt, give it a shot!**

If you’re not sure whether you’re the right fit for a role (and it’s a relatively low-cost action), just apply. If you’re not sure whether you have the skills to do a particular project, time-bound yourself and try it.

(But be wary of doing hard-to-reverse things without careful thought, e.g. being very noisy in the governance space. Replicating papers and posting on your Substack is usually safe.)

**2: Build in public**

One of the best ways to get feedback, signal you’re serious and find collaborators is to tell other people about it! Your ideas will not do you much good sitting in a Google Doc.

Post on LinkedIn, start a Substack, go to[ AI safety events](https://luma.com/bluedotevents?k=c) and talk to people about it. No one is expecting perfection.

**3: Don’t wait for permission**

Never assume someone else is solving the problem you care about.

Even if they are, extra hands are almost always welcome. There’s far more work than people, and they could always use someone with the right skills who’s keen to work on it.

**4: Consider neglectedness.** 

One of the key reasons why AI safety is so important for more people to work on is how few people work in it relative to its importance to the world. Similarly, within AI safety, there are areas and organisations that receive more attention, and others that receive less. 

While the better-trodden paths may have great career capital and be easiest to see, also consider where you have special insight or knowledge that fewer people are working on.

**5: Do things that are [galaxy-brain resistant](https://vitalik.eth.limo/general/2025/11/07/galaxybrain.html)**.

Often it is very easy to rationalise actions (like joining a frontier lab pushing AI capabilities so that you can “be in the room where it happens”). But it can be valuable to take a step back, and consider whether you could use similar arguments to rationalise anything and generally pursue theories of change that are simple, robust, and do not have extremely high downside risk.

---
Most people might find it helpful to explore these broad areas:
## Research

> Implement techniques to make AI safer.

This could involve directly training models, cross-pollinating ideas from other fields or more theoretical work. For example, the concept of model organisms first appeared in biology, and the practice of red-teaming comes from cybersecurity.

Having a deep understanding of ML is incredibly valuable, but those with less ML experience can likely make up for it by[ being the best within your niche](https://www.lesswrong.com/posts/XvN2QQpKTuEzgkZHY/being-the-pareto-best-in-the-world).

This week you could:
- Apply to the [Technical AI Safety Project](https://bluedot.org/courses/technical-ai-safety-project) sprint to replicate and extend an interesting AI safety finding.
- Pick an open problem from orgs like [Open Philanthropy](https://coefficientgiving.org/tais-rfp-research-areas/), [Redwood Research](https://blog.redwoodresearch.org/p/recent-redwood-research-project-proposals), [UK AISI](https://www.aisi.gov.uk/research-agenda), [AISI’s Alignment Team](https://www.alignmentforum.org/posts/tbnw7LbNApvxNLAg8/uk-aisi-s-alignment-team-research-agenda) or [Anthropic](https://alignment.anthropic.com/2025/recommended-directions/) to start working through
- Independently work through technical upskilling programs like[ ARENA](https://www.arena.education/) (or find a collaborator in Slack!)
- Apply to programs like[ MATS](https://www.matsprogram.org/),[ Pivotal](https://www.pivotal-research.org/fellowship),[ LASR](https://www.lasrlabs.org/),[ PIBBS](https://princint.ai/) or the [Anthropic Fellows program](https://alignment.anthropic.com/2024/anthropic-fellows-program/) to gain research experience with a mentor

## Engineering

> Build tools and frameworks that enable researchers to run experiments efficiently.

Some examples include UK AISI’s[ Inspect](https://inspect.aisi.org.uk/) and Anthropic’s[ Petri](https://www.anthropic.com/research/petri-open-source-auditing). Though it may seem “sexier” to work on research problems, engineers have a huge impact by enabling more impactful research by more people. For example, "evals research" is heavily engineering and ops skewed. 

Having a strong software engineering background or even product experience helps develop these tools.

This week you could:
- Apply to the [Technical AI Safety Project](https://bluedot.org/courses/technical-ai-safety-project) sprint to write and improve safety research code.
- Pick an issue on an open-source AI safety tool to resolve
- Build an evaluation tool prototype
- Participate in a hackathon from orgs like[ Apart](https://apartresearch.com/sprints)

## Founding

> Start the initiatives that will make AI systems safer.

There may be orgs already working on the thing you care about that you can join, but sometimes headcount, culture fit, location, and other factors can get in the way. Sometimes, there isn’t anyone working on the particular thing you care about.

Having good ideas and being high agency is incredibly helpful. You can [apply](https://web.miniextensions.com/9UQbEOQ10oTYrqpIOwVS) to our [incubator week](https://bluedot.org/blog/announcing-incubator-week-v2) or other places like [Seldon](https://seldonlab.com/) and [5050](https://www.fiftyyears.com/rfs).


Though working on the models themselves is the most obvious way to contribute, it would be a mistake to think that is the ONLY way to make an impact. AI researchers don’t operate in isolation. They work alongside operations staff, designers, product managers, research managers, recruiters, and subject matter experts.

Many impactful roles exist beyond the categories above. Lean into your unique strengths.

---

**Note**: Technical work is not enough to ensure AI systems are safe. We also need:

- **Governance mechanisms** that enforce safety protocols on all AI developers
- **AI security** to prevent model theft or escape
- **Societal defenses** for harms that slip through (e.g. [biosecurity](https://bluedot.org/courses/biosecurity), cybersecurity)

We’ll cover these areas in other courses.


In this unit, you'll create an action plan for how you could start contributing. This is only the beginning. You're not expected to have a perfect answer right now.

You might find these other courses we offer particularly helpful:
- **[AGI Strategy course](https://bluedot.org/courses/agi-strategy)** for developing your theory of change to steer the trajectory of AI.
- **[Technical AI Safety Project](https://bluedot.org/courses/technical-ai-safety-project)** for building your portfolio or testing your fit as an AI safety researcher or engineer.

You might also want to consider getting free [1-1 career advice](https://80000hours.org/speak-with-us/?int_campaign=primary-navigation) from 80,000 Hours.


Based on everything you've learnt about so far during this course, pick a technical AI safety technique. 

We've also compiled technical research agendas that different orgs are excited about in the optional resources.

To help you prioritise, consider which technique you think would be effective against the threat you developed in Unit 5.


Prioritise a single intervention

- What does success look like with this technique? How does it help make AI safer?
- What's the current status of this technique? Are governments, AI companies or other actors already doing it? If not, why not?
- Which organisations are working on this which you can contribute to or join?

Spend ~1 hour on this.


Do your own research

Now that you've explored the landscape of technical AI safety, it's time to clarify your own path forward.

Start by creating your 1-pager — a concise document that captures what you're looking for and what you bring to the field. This is useful both for your own thinking and as something you can share with potential employers or collaborators.

We don't expect this to be super detailed or a perfect representation. Treat this as a first draft, and something that's _good enough_ to help you get started.

The optional resources offer additional perspective on building a career in AI safety from independent research to landing roles at frontier AI companies. 


Spend ~1 hour on this.

Once you're finished, 1) make the Google Doc shareable, 2) submit it [here](https://bluedot.org/courses/technical-ai-safety/6/7), and 3) share it with your group in Slack!


Fellowships are a common entry point into technical AI safety research. They offer mentorship, structure, and a chance to test your fit before committing to a full-time role. If you're unsure whether you're ready, apply anyway.


Technical understanding is valuable in policy roles. If you're drawn to shaping how AI is governed rather than building safety techniques directly, these fellowships can help you pivot.


Not all paths fit neatly into "technical" or "policy". These fellowships serve specific niches: journalists, interdisciplinary researchers, or those exploring adjacent problems like biosecurity. Worth a look if your background is unusual.


We hope you've had a great time on the course, and we're excited to see how you contribute to making AI go well :) 


Technical AI Safety

How can you contribute to making AI safer?


Start contributing

Why is building safe AI so technically difficult?


The technical challenge with AI

How do we teach AI to behave the way we want?


Training safer models

How do we know if the AI is actually safe?


Detecting danger

What's actually happening inside these black boxes?


Understanding AI

What happens when training and detection fail?


Technical AI Safety

Create your 1-pager

Resources

Exercises