Study Finds Open-Source LLM Agents Currently Unsuitable for Application Security Testing
A new empirical study published on arXiv found that general-purpose open-source large language model (LLM) agents perform poorly compared to established Static Application Security Testing (SAST) tools. Researchers tested three Ollama-hosted open-source models against Bandit, a vetted SAST tool, measuring precision, recall, and false positive rates. The findings suggest that despite growing enthusiasm for agentic AI in cybersecurity, current open-source LLMs are not suitable replacements for specialized security scanning under realistic conditions.
Researchers have published an empirical assessment on arXiv examining whether open-source LLM-based AI agents can substitute for traditional Static Application Security Testing (SAST) tools in cybersecurity workflows. The study evaluated a general-purpose GenAI agent powered by three different open-source models hosted via Ollama, benchmarking their performance against Bandit, an established and vetted SAST tool. Performance was measured across precision, recall, false positive counts, and a composite score designed to capture the interplay among those metrics. Across all evaluated models, the LLM-based agents fell short of Bandit's baseline performance, producing results the authors characterize as unsuitable for real-world SAST scanning tasks. The study contributes to a growing body of research cautioning against premature deployment of general-purpose AI systems in specialized, high-stakes domains such as software security. The authors do not rule out future suitability as models improve, but their current findings present a clear negative result for the open-source LLM-as-SAST-agent use case.
What's missing
The study's own limitations worth noting include: it tested only general-purpose open-source models rather than security-fine-tuned LLMs, which may perform differently; the evaluation is limited to the Bandit tool as a single SAST baseline, which itself has known limitations; the specific programming languages and codebases used in testing are not described in the abstract, limiting generalizability; and the paper has not yet undergone formal peer review, as it is a preprint.
What different sources said
- arXiv cs.AICenter
Can Open-Source LLM Agents Replace Static Application Security Testing Tools? An Empirical Assessment
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.