AI Security Institute
aisi.gov.uk
Actively hiring
AI Governance & Compliance
AI Research Labs
AI Safety & Alignment
As featured in
View 8 jobs
Follow
Overview
News
Technologies
Salaries
Products
People
Growth
Offices
Jobs
Financials
Overview
The AI Security Institute is the United Kingdom government body that tests advanced artificial intelligence models for risks relevant to national security and public safety. Its researchers run pre deployment evaluations together with frontier laboratories, covering cyber offence capability, chemical and biological uplift, autonomous replication and the robustness of model safeguards. The institute publishes its evaluation methodology and the Inspect open source testing framework, and it funds external alignment research through a dedicated grant programme.
News
Anthropic did not submit Mythos 5.1 to UK AI Security Institute
Anthropic released Claude Mythos 5.1 without submitting it to the UK AI Security Institute, the first time it has bypassed the British testing body.
Read more
Report
Can AI agents escape their sandboxes? A benchmark for safely measuring container breakout capabilities AISI Work
We introduce SandboxEscapeBench, the first benchmark to systematically evaluate whether AI agents can break out of their sandboxes, and share some early results.
Read more
Report
Incident Report: unsanctioned agent behaviour during cyber testing AISI Work
During a routine cyber evaluation, AISI identified an incident in which AI agents took sustained, unsanctioned action directed at real people and organisations. We are disclosing what we found, what it means, and the actions now underway.
Read more
Report
UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities AISI Work
Our joint evaluation with CAISI finds Kimi K3 trails leading US frontier closed weight models on cyber capability.
Read more
Report
Cheating behaviour in frontier model evaluations AISI Work
We find cheating behaviour in all of our cyber capability evaluations, and outline the implications as models grow more capable.
Read more
Report
How Far Behind the Frontier are Leading Open Weight Models on Cyber? AISI Work
We evaluated the cyber capabilities of leading open and closed weight AI models, and found that recent open models GLM-5.2 and DeepSeek V4-Pro perform similarly to frontier closed models released 4 to 7 months before them - a narrower gap than the 6 to 10
Read more
Report
How fast is autonomous AI cyber capability advancing? AISI Work
The length of tasks frontier models can autonomously complete in our narrow cyber suite has been doubling every few months. This doubling rate has become faster over time, and recent models exceeded our previous trends.
Read more
Report
Our evaluation of OpenAI's GPT-5.5 cyber capabilities AISI Work
AISI conducted cyber evaluations on OpenAI's GPT-5.5. GPT-5.5 is one of the strongest models we have tested on our cyber tasks and is the second model to solve one of our multi-step cyber-attack simulations end-to-end.
Read more
Report
Our evaluation of Claude Mythos Preview's cyber capabilities AISI Work
We conducted cyber evaluations of Anthropic's Claude Mythos Preview and found continued improvement in capture-the-flag (CTF) challenges and significant improvement on multi-step cyber-attack simulations.
Read more
Report
How do frontier AI agents perform in multi-step cyber-attack scenarios? AISI Work
We tested seven large language models (LLMs) on two custom-built cyber ranges, measuring their ability to execute extended attack sequences in complex environments.
Read more
Report
Show more news
Tech stack
Technologies named in job postings at AI Security Institute over the last 12 months, by layer, with the share of postings for each.
Technologies in use
30
Postings analysed
13
Stack modernity
84/100AI adopter
Languages
Python69%
JavaScript15%
Node JS15%
TypeScript8%
Frameworks & libraries
PyTorch54%
JAX15%
Keras15%
TensorFlow15%
Cloud & platforms
Kubernetes46%
OpenAI38%
AWS31%
Anthropic31%
AI & ML
LLM85%
Claude23%
Sign in to see how your skills match this stack
Sign In
Full AI Security Institute tech stack
30 technologies · by team · pay by technology · changes · similar stacks
Growth
Hiring Momentum
37/100
Slowing
Open positions
9
1 opened / 2 closed in 30 days
Median time-to-fill
127 days
faster than 12% of the market
ATS activity
Every ~4 hours
Today
Hiring Dynamics
0%
Hiring Focus
The percentage next to each role is its share of the company's job openings over the last 90 days; the arrow shows the shift versus the previous period.
Security
70% ▼
AI/ML
20% ▲
Management
10% ▲
Actively hiring for Security (+3 in 30 days)
Activity Timeline
Removed Rest API from stack
Sep 2026
Added Rest API to stack
Sep 2026
Added Machine Learning to stack
Sep 2026
Added Proxmox VE to stack
Sep 2026
Added Claude to stack
Sep 2026
Added vLLM to stack
Sep 2026
Added Terraform to stack
Sep 2026
Added RAG to stack
Sep 2026
Added Docker to stack
Sep 2026
Added Azure to stack
Sep 2026
Added AI Agents to stack
Sep 2026
Added Pre-training to stack
Sep 2026
Offices
Where the company hires and what each office is for
Cities
2
Hiring now
1
Countries
2
Busiest office
London
Loading the map
Headquarters
Builds here
Office with open vacancies
No open vacancies right now Tap a pin for the office
London
United Kingdom
HQ
London
10 open roles 10 posted in 90 days 90% engineering 100% remote $192k Avg Salary
hires for Security, AI/ML, Backend
San Francisco
United States
Regional office
SF
Jobs
Remote (United Kingdom) • London
Fine-tuning
LLM
Anthropic
SFT
Red Teaming
Machine Learning
Apply
Remote (United Kingdom) • London
Fine-tuning
PyTorch
LLM
OpenAI
Anthropic
Pre-training
Red Teaming
Machine Learning
Apply
Remote (United Kingdom) • London
AI Agents
Machine Learning
Apply
Remote (United Kingdom) • London
Python
Claude
LLM
DevOps
Proxmox VE
Apply
Remote (United Kingdom) • London
Python
Claude
AI Agents
DevOps
Proxmox VE
Apply
Remote (United Kingdom) • 1+ year exp • London
Python
Claude
LLM
Anthropic
Post-training
Red Teaming
Machine Learning
Apply
Remote (United Kingdom) • London
Python
JavaScript
Node JS
FastAPI
AI/ML
Fine-tuning
Reinforcement Learning
JAX
AI Agents
TensorFlow
Keras
PyTorch
LLM
Ray
Post-training
Interpretability
Machine Learning
DevOps
SLURM
Docker
Kubernetes
Apply
Remote (United Kingdom) • London
Python
PyTorch
LLM
OpenAI
Machine Learning
DevOps
SLURM
CI/CD
AWS
Kubernetes
Apply