Hacktron AI
2 Tools · 20 Intel
Hacktron AI is a cybersecurity startup building AI agents that autonomously review code, find vulnerabilities and write exploits, bringing continuous, attacker-grade security testing into the software development lifecycle. It was founded by competitive hackers Zayne Zhang, Mohan Pedhapati and Harsh Jaiswal, veterans of DEF CON CTF wins and talks at DEF CON and Black Hat. The company raised $2.9 million in pre-seed funding in 2026, led by Crane Venture Partners.
Is this you? Sign in with X to claim this profile.
Tools
AI code reviewer that finds exploitable vulnerabilities in pull requests and proves them with working exploits.
Free AI-powered pull request security reviews for qualifying open source projects.
Intel
Hacktron's months-long libheif audit found memory-corruption, info-leak and RCE paths through HEIC and AVIF decoding in OpenAI, Slack, Meta, GitHub Enterprise and Next.js. It recommends libheif 1.23.2 or later and sandboxed decoding, and says frontier-model agents sped up exploit development.
Hacktron launches Automation for its security review agent. Rule-based triggers start dynamic exploitability checks, then the agent writes a fix, tests it against the repo's test suite, and opens a pull request. Metrics such as 85% actionable findings are self-reported.
Hacktron audited Cloudflare's AI-built Next.js replacement vinext and validated 24 flaws, 4 critical. Walks through cross-request session leaks, cache key omissions and middleware bypass via double decoding, arguing test-driven AI code leaves the negative space unexamined.
A researcher flags a critical GlobalProtect VPN authentication bypass not limited to PAN-OS like a prior buffer-overflow bug, claiming it already granted VPN access into major corporate networks; full disclosure pending.
Hacktron researchers detail a TOCTOU flaw in Copilot Chat's applyPatchTool: a crafted prompt-injection patch bypasses file-write confirmation, letting an attacker rewrite shell configs or .git/config for RCE via agent mode.
Researchers chained a libheif heap overflow reachable via OpenAI's help forum with an SSO flaw to hijack employee ChatGPT/Codex accounts, proving internal repo access with a harmless PR before responsible disclosure and a $6,500 bounty.
GitHub Enterprise Server shipped a regex flaw in OAuth redirect-URI validation that let attackers craft authorization links redirecting a victim's OAuth code to an attacker-controlled domain, enabling repository takeover; now patched.
Hacktron launches Whitebox, an AI-assisted white-box security assessment that reasons over source code to catch business-logic and auth-boundary bugs black-box testing misses, with human-reviewed, audit-ready reports.
Researchers chained a libheif heap overflow from a HEIF upload into RCE, then an OpenAI SSO flaw, to take over employee ChatGPT and Codex accounts and push a proof-of-concept PR into OpenAI's internal monorepo.
Hacktron shows a bypass of VS Code's patch for prompt injection in Copilot agent mode. A time-of-check to time-of-use gap in the apply-patch tool lets a crafted GitHub issue overwrite sensitive files and leak the developer's GitHub token. The post includes code paths and a proof of concept.
A critical remote code execution vulnerability was found in self-hosted Next.js deployments; Vercel-managed hosts are unaffected, with a full technical write-up to follow.
A pre-authentication remote code execution bug via SQL injection (CVE-2026-63030) hits WordPress core 6.9.0-6.9.4 and 7.0.0-7.0.1, while 6.8.5 and earlier remain unaffected.
An active supply-chain compromise is spreading through npm packages published under the Immobiliare organization; teams should audit dependency trees, remove affected packages, and rotate exposed secrets.
Hacktron AI is flagging a live, actively circulating vulnerability in Astro servers below version 7.2.8. No public PoC yet, but Astro operators should patch now.
Rocket.Chat put Hacktron's AI reviewer on every pull request and within 20 days caught 17 real vulnerabilities, including a critical account-takeover flaw (CVE-2026-55666) and a token-replay bug (CVE-2026-55759).
Hacktron ships an MCP server that lets Claude Code, Codex, or Cursor take a pull-request URL, pull the relevant security findings, write the fix, and mark the issue resolved automatically inside a normal agent session.
Three researchers used AI models, including Claude Opus 5, to rediscover an unpatched image-processing bug lacking a CVE, reported it to OpenAI for a $6,500 bounty, and fixed in 14 hours, warning other companies on the same stack remain exposed.
Hacktron opens its AI pull-request security reviewer free to qualifying open-source maintainers, citing a drop in median time-to-exploit to roughly -18 hours; it posts inline findings, learns project patterns, and auto-resolves fixed issues.
Hacktron's automated review found a high-severity NoSQL injection in Novu's codebase by tracing an unvalidated ID through an ingest controller, a type cast, and a sink service across three files.
Hacktron explains why it treats public pull requests as security events: full vulnerability details in a public PR comment tip off attackers, so on public repos it shows only a link, keeping specifics visible to maintainers only.