@goodfellow_ian: In Aisle's first year, they've shown that the system built around an LLM is highly important, and even small, cheap fas…
Summary
Ian Goodfellow highlights AISLE's first year results: an AI security research lab that discovered 500+ zero-day vulnerabilities with CVEs across 100+ open-source codebases, rivaling frontier AI labs even with small, cheap, fast models. The key insight is that the system built around an LLM matters as much as the model itself.
View Cached Full Text
Cached at: 10/17/26, 02:47 AM
In Aisle’s first year, they’ve shown that the system built around an LLM is highly important, and even small, cheap fast models can be boosted to do valuable security work
Stanislav Fort (@stanislavfort): My AI security research lab AISLE discovered 500+ zero-days with CVEs across 100+ codebases in a year => ~1% of 2026’s open-source CVEs globally from a single team! On hardened codebases like curl, we’ve repeatedly matched or outperformed frontier AI labs in zero-day discovery
Similar Articles
@stanislavfort: My AI security research lab AISLE discovered 500+ zero-days with CVEs across 100+ codebases in a year => ~1% of 2026's …
Stanislav Fort's AI security research lab AISLE claims to have discovered 500+ zero-day vulnerabilities with CVEs across 100+ codebases in a year, roughly 1% of 2026's global open-source CVEs, while matching or outperforming frontier AI labs on hardened codebases like curl.
We have a year to fix security everywhere
The article discusses the release of GLM 5.3-flash, an open-weight AI model that is cheap and capable, raising concerns about potential misuse for hacking. It calls for urgent action to fix security vulnerabilities across the tech industry using frontier LLMs.
@hetmehtaa: Local AI for Penetration Testing & Research https://projectblack.io/blog/local-ai-for-cyber-security/…
A blog post benchmarks four approaches (Semgrep, GLM 5.1 with Strix, cloud SOTA with code review skill, and local AI with a custom harness) for finding a known LFI vulnerability in PHPIPAM, finding that the local AI harness with a tailored approach outperforms the others.
A fundamental flaw leaves LLMs strikingly vulnerable to attack
Researchers present a paper at ICML arguing that a fundamental flaw in how LLMs identify instructions makes them impossible to fully secure against attacks, demonstrating successful exploits against models from OpenAI, Anthropic, Alibaba, and DeepSeek.
Gate AI: LLM Security Benchmark Evaluation Methodology and Results
This paper presents an evaluation methodology for LLM security detectors that addresses systematic weaknesses like per-dataset threshold tuning and undisclosed operating points. The framework uses cross-validation across 16 benchmarks, selects a single global operating point, and includes multiple diagnostics for generalization.