Our view on the emerging challenges facing frontier AI organisations, the technical practices most urgently needed to meet them, and how we will work with industry and others to develop them.
How do we test AI systems for dangerous capabilities without risking real-world harm? The more capable models become, the harder it is to safely evaluate them. When an agent can execute arbitrary code ...
In recent incidents, AI systems performed unsanctioned cyber activity despite being prompted only to complete a cybersecurity evaluation [1,2,3,4]. This includes AI systems engaging in supply-chain ...
An update on the security changes we have recently made to our frontier AI evaluations, and the work that remains.
We introduce Transect, an open-source tool that helps evaluators follow an agent’s work, identify what to investigate closely, and check their interpretation against the transcript.
Abby D’Cruz, Alastair Pearson, Alex Anwyl-Irvine, Alexandra Souly, Aliya Ahmad, Anna Gausen, Barnaby Perkes, Ben Millwood, Catherine Fist, Christopher Summerfield ...
The widespread adoption of AI systems for information seeking has raised concerns about their potential impact on democracy. Many people worry that biased or ‘hallucinated’ outputs could misinform ...
Several leading AI companies have the stated goal of building AI systems that match or surpass humans across most cognitive domains. Such systems could have transformative effects, such as increasing ...
AI systems are increasingly being used internally within frontier AI companies to assist with research and development, including AI safety work. However, if these systems behave in misaligned or ...
Alongside Irregular, we found evidence demonstrating that evaluators need to use large token budgets to understand the cyber capabilities of recent Large Language Models (LLMs). Are AI evaluations ...
Identifying dangerous AI capabilities requires rigorous evaluation of models at the upper limit of their abilities—a process that is far from straightforward. There are many ways to enhance the ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results