November 11, 2026

Firewall Flaws, Token Theft, and 40K Exposed Secrets

Firewall Flaws, Token Theft, and 40K Exposed Secrets

The short version: a lot of doors are still open

The week ending October 29, 2026 brought a heavy mix of patching emergencies and credential disasters. If you’re running Fortinet gear or Microsoft Authenticator in a business environment, those are the two things that need attention before anything else.

For small and mid-size businesses in NJ and NYC, the practical risk here isn’t abstract. Stolen credentials from FortiGate configs, a flaw that let attackers grab Microsoft work account tokens without any special permissions, and fake AI bots systematically scraping .env files from company servers. These aren’t theoretical attack chains. They’re happening on a schedule, around the clock.

The Fortinet situation is worse than one patch cycle

Fortinet released a coordinated set of security updates for FortiOS, FortiManager, and FortiAnalyzer in late October 2026. The details are being held until November 10, 2026, which is standard coordinated disclosure, but the urgency is real. Fortinet devices have become a preferred entry point for attackers, and the FortiBleed campaign makes that concrete.

Hudson Rock’s follow-up investigation into FortiBleed uncovered 975 newly identified domains and 40,573 exposed secrets pulled from 1,703 FortiGate source files. That includes 37,076 local-user records and 1,576 Active Directory LDAP bind records. The configuration data dates to around February 2026. This is now being called FortiBleed 2.0.

If you have Fortinet hardware and haven’t verified your firmware is current, that’s the first call to make. LDAP bind credentials showing up in leaked configs means attackers potentially have a path into Active Directory, which touches everything else in your environment.

Microsoft Authenticator had a zero-permission token theft flaw

CVE-2026-80097, nicknamed AuthMask, is now patched, but it’s worth understanding what it allowed. An unprivileged Android app with zero permissions could impersonate any Microsoft first-party app and convince the Authenticator broker to hand over authentication tokens for that app. No elevated access required. A user just had to complete an authentication flow in what looked like a legitimate app.

Microsoft confirmed that successful exploitation could allow an attacker to obtain authentication tokens and access resources as the affected user, including work accounts. If your team uses Microsoft 365 and authenticates via Android devices, make sure Authenticator is updated. This is exactly the kind of quiet flaw that doesn’t make headlines until after someone gets hit.

We’ve written about how AI-assisted vulnerability discovery is changing the pace of patch releases. AI is finding more bugs, and your patches genuinely can’t wait the way they used to.

Fake AI bots are scraping your servers for secrets

Cequence tracked a single attacker running 32 fake AI bot identities, including impersonations of ChatGPT, Claude, and Perplexity, to scan for credentials. The scanner worked through roughly 90 specific file paths targeting .env configuration files, cloud credential files, private SSH keys, and server environment data. It hit organizations in financial services, retail, telecommunications, energy, and software from at least late September through late October 2026.

This matters if you have any public-facing web infrastructure. Misconfigured or forgotten credential files are a real exposure point, and the scanning is automated and constant.

There’s a broader pattern worth understanding here. AI tools creating new credential and code exposure risks is something we’ve covered before, and this campaign fits that same pattern of attackers borrowing AI’s trusted reputation to do something malicious.

One more thing: don’t share banking credentials with AI agents

Bank of America sent customers a direct warning in late October 2026: never share your username and password with an AI-enabled tool or agent that offers to act on your behalf. Worth passing along to your team. The social engineering angle on AI tools is getting more sophisticated.

What to do this week

  • Apply the Fortinet updates for FortiOS, FortiManager, and FortiAnalyzer immediately. Don’t wait for November 10.
  • Update Microsoft Authenticator on all Android devices used for work accounts.
  • Audit any public-facing servers for exposed .env or credential files.
  • Check whether your Active Directory credentials appear in any known breach data, given the FortiBleed 2.0 findings.

These aren’t complex fixes individually, but they pile up fast if you’re managing them without a system. If you want a second set of eyes on your current patch posture and credential exposure, booking an IT assessment with Exine is a practical starting point.

FAQ

What is the FortiBleed 2.0 breach and does it affect my business?

FortiBleed 2.0 is a credential harvesting campaign involving leaked FortiGate configuration files. Hudson Rock found 40,573 exposed secrets in data dating to around February 2026. If your organization uses Fortinet devices, check for firmware updates and verify your Active Directory credentials haven’t been compromised.

Is Microsoft Authenticator safe to use after the CVE-2026-80097 flaw?

Yes, as long as it’s updated. CVE-2026-80097 has been patched. The flaw allowed a zero-permission Android app to steal authentication tokens. Update the app on all work devices and you’re covered. Leaving it unpatched is the actual risk.

How do attackers use fake AI bot identities to steal credentials?

They impersonate known AI crawlers like ChatGPT or Claude to avoid detection, then systematically request files that commonly store secrets, like .env files and SSH keys. Cequence tracked one actor using about 32 fake bot identities hitting targets across multiple industries through late October 2026.

Should employees share passwords with AI tools that manage tasks for them?

No. Bank of America issued a direct warning in October 2026 advising customers never to share usernames or passwords with AI agents or tools, even ones that claim to act on your behalf. Treat AI tools the same way you’d treat any third-party service: don’t hand over core credentials.

Tomasz Sobolewski, founder of Exine LLC
About the author
Tomasz Sobolewski
Founder of Exine LLC. Hands-on IT, cybersecurity and backup for growing New Jersey businesses, with 15+ years in the field. The kind of support that knows your systems and picks up the phone.