AlphaProof Code is an AI-assisted framework for formally verifying complex software: it uses mathematical proofs to establish that code satisfies specified security properties. Google DeepMind researchers applied it to the implementation of Signal’s cryptographic protocol, which protects conversations in Signal and Google Messages. Unlike verification that checks only a protocol’s design, this effort traced properties through compiled code to the software running on devices, addressing vulnerabilities that can arise when implementation diverges from design. Researchers and AI models first translated security goals into Lean specifications, which human experts reviewed. AI agents then generated more than six million lines of proof code, while Lean independently checked each proof.
The team says the work is the first formal verification of a widely deployed cryptographic protocol’s implementation, and the first use of AI to verify such a protocol at this depth. The process took four weeks, compared with the years such work could historically require; agents also directed human reviewers to roughly one percent of the codebase for closer inspection. As a test, researchers inserted two previously fixed bugs into an isolated Signal codebase, and the system proved that each buggy version was insecure. The verified properties include protecting messages before and after a device is temporarily compromised. The researchers are releasing verification artifacts and argue that the approach could extend rigorous assurance to other critical software, including medical devices and power-grid systems, while requiring further research to scale.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers tested 16 LLMs on all 55 TIS-100 puzzles, giving each one attempt and time to refine its code. The models found 133 new performance tradeoffs across 18 puzzles and set three outright cycle records, though human players improved on most solutions within days. (tis-100.wincon.dev)
Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent harm caused by their models. The government is also weighing an AI safety watchdog with power to impose significant fines, following a hack of a Medicare statistics portal involving OpenAI. (abc.net.au)
AI is increasingly becoming a gateway through which people find information. Audiences still resist having AI produce their news. (niemanlab.org)
Nvidia holds a central position in the AI industry, making disruption to the company potentially catastrophic for the wider ecosystem. The race for advanced AI is also intensifying economic and geopolitical conflict, while concentrated corporate power and infrastructure face growing political resistance. (france24.com)
LLM inference can require reading about 26 GB of model weights to generate each token from a 13-billion-parameter model. The explanation shows how model size drives memory bandwidth demands during text generation. (youtube.com)
Deletion Tuning matches full fine-tuning across three LLM post-training methods by zeroing fewer than 0.05% of model weights. Its compact masks also preserve prior-task accuracy better than full fine-tuning, LoRA, and random sparse tuning. (arxiv.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.