Agent Builder by AirtopWeb agents that heal themselves.
Promoted
Aria-Sec
AI security that earns your trust before it acts
16 followers
AI security that earns your trust before it acts
16 followers
I’m building ARIA, an AI security analyst that investigates threats and can take action within controls set by the customer. Today, companies either overwhelm analysts with alerts or ask them to trust automation. ARIA starts by requesting approval, verifies every outcome, and earns narrowly scoped autonomy over time - it cannot promote itself. I taught myself software development this year and built the working desktop and web platform, backend, security connectors, and governance system myself.
Congrats on shipping this solo. The scoped autonomy approach sounds thoughtful. One thing that would help build trust: a clear audit log export that shows every action ARIA took, the approval chain, and what data it accessed. SOC 2 reviewers will ask for it anyway, and having it visible in-app would make the "verify every outcome" promise tangible instead of just a claim.
@navflow Thank you - you’re right about the trust value here. ARIA already maintains an in-app, tenant-scoped audit ledger covering actions, actors, approvals, outcomes, and governance changes.
The gap is packaging that evidence into a clear, exportable chain - including the specific systems and data each operation accessed - for operators and SOC 2 reviewers. That’s a useful distinction, and I’m adding the unified export experience to the roadmap.
Really cool approach with the trust-building autonomy model. One thing that would help teams adopt this faster: a sandbox mode where ARIA investigates a copy of recent alerts in parallel with human analysts, then compares findings side by side. That way you can show measurable accuracy and false positive rates before anyone grants it any action permissions.
@sian_tate That’s a great point. ARIA already supports a version of this in AI SPM: users can investigate findings in a sandboxed environment, with human approval required before any action is taken.
Right now, though, that capability is limited to the AI SPM panel. A universal sandbox mode—allowing ARIA to work alongside analysts across all functions and compare results side by side—is definitely something I need to build.
Pazi
Such a promising launch—congrats on Aria-Sec! AI security that earns your trust before it acts is a clever idea. Wishing you a strong Product Hunt day! 🚀
@zvonimir_sabljic1 Thank you, appreciate the comment. I applied to Y combinator with the idea so lets see what happens