Future of Life Institute
A nonprofit working on extreme risks from transformative technology; it built secure-hardware proofs of concept for AI governance and partners on a network-logger prototype.
futureoflife.org · also called FLI
FLI describes itself as a non-profit whose mission is to steer transformative technology towards benefiting life and away from extreme, large-scale risks 1. Its verification work includes:
- With Mithril Security, FLI built a secure-hardware proof of concept on Intel SGX and the BlindAI framework in which a model owner leases weights to an untrusted party 2. It was designed to keep the weights inaccessible to the borrower, report how much computing is done and provide an off-switch 2. FLI calls it "not necessarily deployable as is", citing performance and hardware attacks that need mitigation 2. See TEE remote attestation for AI workloads.
- A 2024 FLI post with Mithril Security reports AICert, a proof of concept in which Trusted Platform Modules bind a model's weights to the code and data used to train it 3. The authors state that it covers fine-tuning only, has not been audited by a third party and does not detect poisoned models or datasets 3.
- The AI Futures Project's verification page lists FLI and SASH as building a confidential network logger prototype 4, and SASH names FLI as an early partner in that work 5. See SASH confidential network logger.
On this page
Related records
Mechanisms and implementations whose records cite or describe this organization's work.
- Trusted execution environments (TEEs) in CPUs and GPUs sign reports of loaded software, so a remote party can check which code ran an AI workload.
- An open-source prototype that routes a facility's inference traffic through a logger and re-runs requests on a separate cluster to check it serves inference.
Publications
Sources this organization authored or published.
- BFuture of Life Institute: Global Institutions Governing AI. Future of Life Institute. Source recordCited by Future of Life Institute
- CA. Aguirre & R. Millet (2024). Verifiable Training of AI Models. Future of Life Institute. Source recordCited by Future of Life Institute
- CFuture of Life Institute (2023). Exploration of secure hardware solutions for safe AI deployment. Future of Life Institute. Source recordCited by TEE remote attestation for AI workloads; Future of Life Institute
Mentioned in
Sources
- BFuture of Life Institute: Global Institutions Governing AI. Future of Life Institute. Source recordSupports: self-description as independent nonprofit and mission to steer transformative technology away from extreme risks
- CFuture of Life Institute (2023). Exploration of secure hardware solutions for safe AI deployment. Future of Life Institute. Source recordSupports: Intel SGX proof of concept with Mithril Security: design aims and stated limits
- CA. Aguirre & R. Millet (2024). Verifiable Training of AI Models. Future of Life Institute. Source recordSupports: AICert proof of concept with Mithril Security: TPM-based binding of weights to training inputs; stated limits
- CAI Futures Project (2026). Get Involved in Verification. AI 2040. Source recordSupports: listed with SASH as building a confidential network logger prototype
- CSingapore AI Safety Hub (SASH) (2026). Internationalising AI Verification. SASH blog. Source recordSupports: named by SASH as an early partner