Can AI coding agents rebuild biological AI tools from a paper?
AISI's ReproBAIT evaluation finds coding agents can rebuild simple biological AI tools from limited public info within 4 hours and under $20, undermining withheld-code access controls.
Research
UK AI Security Institute
The UK government’s AI Security Institute: evaluations, red-teaming and research on frontier AI risks. aisi.gov.uk
Focus All
Topic Security & misuse
Source UK AISI · 16
Date Any time
1 item
1 item
How Far Behind the Frontier are Leading Open Weight Models on Cyber?
AISI measures cyber capabilities of open and closed-weight models, finding leading open models now trail frontier closed models by 4-7 months, and discusses misuse risks of open release.
1 item
Our evaluation of OpenAI's GPT-5.5 cyber capabilities
UK AISI reports cyber evaluations of GPT-5.5, finding it is the second model to complete their multi-step corporate network attack simulation end-to-end, suggesting a broader frontier trend.
1 item
AISI Frontier AI Trends Report (2025)
UK AISI's first public report synthesising two years of frontier model testing, tracking trends in dangerous capabilities across national-security and public-safety domains for policymakers and the public.
1 item
Our approach to tackling AI-generated child sexual abuse material
UK AISI and Thorn publish a safety protocol for AI developers to prevent AI-generated CSAM, alongside legislation allowing authorised testing of whether models can generate such material.
1 item
Open technical problems in open-weight AI model risk management
Lists 16 open technical challenges for managing risks of open-weight frontier models across training data, algorithms, evaluations, deployment, and ecosystem monitoring, noting limited safety tooling.
1 item
Examining backdoor data poisoning at scale
Largest data-poisoning study finds as few as 250 documents can backdoor LLMs across 600M–13B sizes, implying poisoning attacks are more feasible than assumed.
1 item
From bugs to bypasses: adapting vulnerability disclosure for AI safeguards
Explores whether cybersecurity vulnerability-disclosure practices can help manage bypasses of generative AI safeguards, in collaboration with the UK NCSC.
1 item
Managing risks from increasingly capable open-weight AI systems
UK AISI post on managing risks from capable open-weight models, covering why released weights let safeguards be cheaply removed, existing mitigation methods, and open problems.
1 item
An example safety case for safeguards against misuse
Proposes an end-to-end safety case showing that misuse safeguards reduce AI assistant misuse risk to low levels, combining red-teaming effort estimates with a quantitative uplift model.
1 item
Making safeguard evaluations actionable
Proposes how to make misuse-safeguard evaluations decision-relevant by setting metrics and thresholds, and presents an example safety case linking safeguard evaluation results to deployment decisions.
1 item
Strengthening AI resilience
UK AISI announces 20 seed grants for independent research on safeguarding critical infrastructure and societal systems against AI-related risks, spanning misinformation, worker impacts and decision safety.
1 item
How we’re addressing the gap between AI capabilities and mitigations
UK AISI outlines its technical research agenda on safeguards against AI misuse and loss of control, covering the widening gap between capabilities and mitigations, plus hiring and grants.
1 item
Pre-Deployment evaluation of OpenAI’s o1 model
UK and US AI Safety Institutes jointly ran pre-deployment tests of OpenAI's o1 across cyber, biological, and AI-development capabilities, comparing it against reference models and sharing findings before release.
1 item
Safety case template for frontier AI: A cyber inability argument
Proposes a safety case template using CAE structured arguments to show frontier AI models lack unacceptable offensive cyber capabilities, linking risk models, proxy tasks and evaluation results.
1 item
Safety case template for ‘inability’ arguments
Proposes a safety case template for arguing that an AI system lacks offensive cyber capabilities, combining existing techniques as a proof of concept for frontier AI assurance.