Loupe is a tool that uses artificial intelligence (AI) to detect vulnerabilities in open source Bitcoin project code, and was announced by Spiral and Block on May 12th.
According to the announcement, the tool combines large-scale language models (LLM), AI systems such as GPT, Grok, and Claude, which can read and analyze large amounts of code. In addition to LLM, Loupe also has its own software layer optimized for Bitcoin projects.
loupe central filter I’m just reporting bugs backed by demonstrable test casesto avoid low-quality reports that overload developers.
The Spiral Team warned: Hackers and attackers are already using AI to identify flaws in Bitcoin codeOn the other hand, those who develop and maintain these projects on their own don’t have the time and tools to do the same. Loupe aims to fill that gap.
Spiral and Block’s stated goal is to transfer control of Loupe to the developers of each project once the tool proves useful. The plan includes three stages: internal scans with responsible disclosure, expansion to more repositories, and finally transfer of control to each community.
Release without evidence or cost to developers
According to a statement posted on Spiral Substack, eight ecosystem repositories are participating in the initial testing phase, including Bitcoin Core, BDK, LDK, rust-bitcoin, Cashu, Blockstream Jade, bitcoinj, and SRI.
Release of loupe No vulnerabilities or flaws foundHowever, the reports published by Spiral include a commitment to responsibly disclose detected errors to each project’s team.
Loupe, on the other hand, can be used independently by any team of developers. However, you will have to pay for access to the AI model that the tool uses for scanningBecause each query to these systems incurs a fee. Early stage project scanning is performed by Spiral and Block.
AI: Tools of defense and attack
The launch of Loupe comes amid the increasing use of AI to both attack and protect cryptocurrency infrastructure.
As reported by CriptoNoticias on May 11, Binance assured that its AI security tools have prevented US$10.53 billion in potential losses from Q1 2025 to Q2 2026. The exchange said 57% of its anti-fraud controls are already powered by AI models, which should reduce the number of card frauds by 60% to 70% compared to the industry average.
However, Binance Research says that AI is currently Exploiting vulnerabilities is twice as effective as detecting vulnerabilitiesAI-powered fraud can be up to 4.5 times more profitable than traditional fraud.
Additionally, the Bybit exchange reported intercepting US$300 million in fraudulent withdrawals in the fourth quarter of 2025. Uses AI-based detection system. Throughout 2025, our systems blocked more than 3 million account takeover attempts.
In this scenario, where AI is already acting as both an offensive and defensive tool, Loupe aims to fill a specific gap. This means that open source Bitcoin projects, with limited resources, can detect vulnerabilities at the same rate as those trying to exploit them.
(Tag Translation) Bitcoin (BTC)
