.svg)
.svg){kind=icon}
Dear Apart Community,
Welcome to our newsletter - Apart News!
At Apart Research there is so much brilliant research, great events, and countless community updates to share.
In this week's edition of Apart News we have two *NEW* papers, the EA Forum's Donation Election, and news from another two of our hackathons this week, with Howard University and Goodfire.
Move Aside US Elections
The EA Forum’s Donation Election (2024) has begun! If you wanted to vote for us to receive some additional funding, all you have to do is sign in to your EA Forum account and vote here.
In just two years, Apart Research has established itself as a unique and efficient part of the AI safety ecosystem. Our research output includes 13 peer-reviewed papers published since 2023 at top venues including NeurIPS, ICLR, ACL, and EMNLP, with six main conference papers and nine workshop acceptances.
Our work has been cited by OpenAI’s Superalignment team, and our team members have contributed to significant publications like Anthropic’s “Sleeper Agents” paper.
With this track record, we’re able to capitalize on our position as an AI safety lab and mobilize our work to impactful frontiers of technical work in governance, research methodology, and AI control. Given current AGI development timelines, the need to scale and improve safety research is urgent.
In our view, Apart seems like one of the better investments to reduce AI risk. If you agree please vote for us here.
*NEW Paper*: CryptoFormalEval
Paper authored by Cristian Curaba, Denis D’Ambrosi, Alessandro Minisini, & Natalia Pérez-Campanero Antolín.
Apart Research’s newest benchmark - CryptoFormalEval - offers a systematic way to evaluate how well Large Language Models (LLMs) can identify vulnerabilities in cryptographic protocols.
By pairing LLMs with Tamarin, a specialized theorem prover, we have created an automated system that could transform how we approach protocol security verification. Read the full paper here.
In our increasingly connected world, cryptographic protocols form the backbone of digital security, ensuring the confidentiality and integrity of our systems of communications, from SSH securing internet communications to OAuth enabling passwordless authentication.
Cryptographic protocols are the practical instruments that protect our digital exchanges. Yet, the rush to deploy these protocols can sometimes outpace the necessary comprehensive verification, leaving potential security gaps.
While formal verification methods exist, they are typically manual, expert-driven, and not scalable—challenges that CryptoFormalEval seeks to address by leveraging the potential of LLMs.
Read our blog here to learn more!
Another *NEW Paper*: CyberSecEvals
This paper is authored by Suhas Hariharan, Zainab Ali Majid, Jaime Raldua Veuthey, Jacob Haimes.
The risk posed by cyber-offensive capabilities of AI agents has been consistently referenced - by the National Cyber Security Centre, AI Safety Institute, and frontier labs - as a critical domain to monitor.
A key development in assessing the potential impact of AI agents in the cybersecurity space is the work carried out by Meta, through their CyberSecEval approach (CyberSecEval, CyberSecEval 2, CyberSecEval 3). While this work is a useful contribution to a nascent field, there are features that limit its utility.
Exploring the insecure code detection part of Meta’s methodology, detailed in their first paper, we focus on the limitations - using our exploration as a test case for LLM-assisted benchmark analysis.
Goodfire-Apart Hackathon
Our mechanistic interpretability Hackathon kicks off today with Goodfire AI. Participants will deep dive into mechanistic interpretability and feature manipulation. Goodfire is bringing unprecedented access to their state-of-the-art tools for understanding and steering AI behavior.
If an idea is promising, participants may be asked to join our brand new Apart Lab Studio to continue the development of their research project.
Howard University's AI Policy Hackathon
We had some really incredible submissions for yet another AI Policy Hackathon in Washington D.C., at Howard University this week. They also had a musical band for mid-hack entertainment!
Have a look here for some pictures from the event.
Opportunities
- Never miss a Hackathon by keeping up to date here.
- Final call for applications for the UK's AI Safety Institute Fast Grants programme is closing soon! Don’t miss the chance to secure up to £200k in funding to advance critical AI Safety research.
Have a great week and let’s keep working towards safe and beneficial AI.
