{
    "success": true,
    "data": {
        "id": 1777479,
        "msgid": "cisco-unveils-agentic-ai-framework-to-address-ais-bug-detection-errors-1780313131",
        "date": "2026-06-01 16:04:00",
        "title": "Cisco Unveils Agentic AI Framework to Address AI's Bug Detection Errors",
        "author": "Wahyunanda Kusuma Pertiwi",
        "source": "KOMPAS",
        "tags": "",
        "topic": "Technology",
        "summary": "Cisco has launched an open-source framework called Foundry Security Spec to enhance AI-based security evaluation systems. The solution tackles inaccuracies in AI-generated bug detection, which often produce unstructured results and false positives. It aims to help organisations focus on verifiable security findings rather than overwhelming notifications.",
        "content": "<p>US-based computer networking and cybersecurity firm Cisco has\nintroduced Foundry Security Spec. Foundry Security Spec is a proven\nframework for building agentic AI-based security evaluation systems,\nreleased as open-source software. This blueprint is designed to assist\norganisations in building AI-powered software security evaluation\nsystems. The solution addresses a growing challenge in cybersecurity as\nAI usage increases: inaccurate or unverifiable bug findings. Generative\nAI models and large language models (LLMs) are increasingly used to\nidentify software vulnerabilities. However, Cisco notes current\napproaches have significant shortcomings. Results often consist of\nunstructured findings, mixing valid analysis with false positives, with\nno clear way to confirm complete coverage or when the evaluation process\nis truly complete. Cisco states this makes AI appear more as a\ntechnological showcase than a reliable security evaluation system for\ncorporate security teams and auditors. \u201cBy sharing our learned\nknowledge, we aim to help the security defender community move faster\nand smarter,\u201d said Omar Santos, Cisco\u2019s Distinguished Engineer for AI\nSecurity Engineering in S&amp;TO, in an official statement.\n\u201cOrganisations will no longer be bogged down by numerous notifications\nbut can focus on verifiable security findings with real impact,\u201d he\nadded. Although announced as an open-source project, Foundry Security\nSpec is not a ready-to-use security application or scanner. Cisco\nexplained that Foundry is a reference specification or blueprint\norganisations can use to build their own AI-based security evaluation\nsystems tailored to their needs. Simply put, while ready-to-use security\ntools are like finished houses, Foundry Security Spec is akin to a\ndetailed architectural blueprint showing how to build them. Being\nmodel-agnostic and stack-agnostic, the specification can be used with\nvarious AI models and software infrastructures, meaning organisations do\nnot need to adopt a specific AI model. Cisco designed Foundry to work\nwith GitHub Spec-Kit, a widely adopted industry specification-based\ndevelopment workflow.<\/p>",
        "url": "https:\/\/jawawa.id\/newsitem\/cisco-unveils-agentic-ai-framework-to-address-ais-bug-detection-errors-1780313131",
        "image": ""
    },
    "sponsor": "Okusi Associates",
    "sponsor_url": "https:\/\/okusiassociates.com"
}