If you don't manage that server yourself, the real job is finding out whether your phone provider runs FreeSWITCH and getting a written answer that it is fully up to date.

This guide explains the problem in plain words, gives you a short list of questions to send your VoIP provider today, and shows how to keep missed-call handling and AI answering safe while you wait for answers.

What happened, in plain words

FreeSWITCH is open-source software that many phone systems are built on. It sits behind the scenes: it connects calls, runs phone menus, records voicemail, and passes calls to other tools such as an AI voice agent or a CRM. You may never see the name. Your phone bill will show a provider's brand, not the engine underneath.

The vendor's advisory says an inbound SIP MESSAGE can select the "api" chat setting and run FreeSWITCH API commands. In other words, according to the vendor advisory, an inbound SIP MESSAGE can run FreeSWITCH API commands on your phone server.

How bad is that? The vendor puts the impact as command execution with the privileges of the FreeSWITCH process: arbitrary operating-system commands, disclosure of configuration and credentials, and full call control. In everyday terms, an attacker could:

The vendor states that exploitation of this path has been observed in the wild. The advisory, GHSA-mg63-h38w-8p9p, was published on Oct 28, 2026. CISA then added CVE-2026-116779, the SignalWire FreeSwitch code injection vulnerability, to its known exploited vulnerabilities catalog.

9.8
CVSS severity score
1.11.3
FreeSWITCH version with the fix
1.11.2
Affected: this version and earlier

Security reporters at eSecurity Planet note that FreeSWITCH 1.11.2 and earlier contain the flaw in how SIP MESSAGE is handled, and that the vulnerability carries a CVSS score of 9.8. That score sits at the top of the "critical" range. The Hacker News lists the flaw with a CVSS score of 9.8.

Why a phone server is a bigger deal than it sounds

Think of your business phone system as an always-online computer that sits in the middle of the most sensitive moments in your business:

Someone who controls the call flow can do more than eavesdrop. They could reroute your main line to a number they control and answer as your business. They could place expensive calls through your carrier account. They could pull the saved passwords and move from the phone server into email or your scheduling tool.

That last point matters because fraud carried out over phones and the internet is already big business. A Help Net Security report on cybercrime losses notes that cyber-enabled fraud totaled $17.7 billion, representing 85% of all reported financial damage. A hijacked business phone line is a ready-made tool for exactly that kind of fraud, because people trust a call that comes from a number they recognize.

Does your office run on FreeSWITCH? How to find out

You don't need to log into anything technical. Work through these in order and stop when you have an answer.

1. Identify who actually runs your phones

Write down the company you pay for phones, the company that installed them (often a local IT or telecom reseller), and any add-on services: call-center software, an after-hours answering service, call tracking for ads, and any AI receptionist or voice agent. Each of these may run its own phone server.

2. Sort your setup into one of three types

Your setupWho has to patchWhat you do
Hosted cloud phones (you pay a monthly per-user fee, no server in your office)The providerAsk the questions below; get the answer in writing
A phone server in your office or your own cloud account, set up by a reseller or IT firmYour IT firm or reseller, on your behalfAsk them to check the version today and schedule the update
An AI receptionist or voice bot connected to your phone numberThe AI vendor, plus whoever connects it to your phonesAsk both parties; the connection between them counts too

You won't reliably spot FreeSWITCH from an invoice or a desk phone. Ask whoever operates the server; their answer is the one to rely on.

Tip

Ask in writing, by email. A written "we are not affected" or "we patched on this date" is something you can file with your security records, and it puts the responsibility where it belongs.

Questions to ask your VoIP provider this week

Copy these into an email to your phone provider, your IT firm, and any AI receptionist vendor. Mention the CVE identifier by name, so the request reaches the right person.

  1. Does any part of the service we use run on FreeSWITCH? This includes call routing, voicemail, call recording, conferencing, and any AI voice or call-center tools.
  2. If yes, which version is running now, and when was it updated? The answer you want is 1.11.3 or later.
  3. If it isn't updated yet, has inbound chat been turned off? Ask whether they have applied the vendor's recommended workaround for systems that can't be updated right away.
  4. Did you rely on firewall or access-list rules alone? SignalWire says the inbound ACL does not cover MESSAGE, so tightening it does not take a profile out of scope.
  5. Have you checked for signs that someone already got in? CISA's guidance for this entry also points to its forensics triage requirements. Put simply, ask whether anyone got in before the fix, not just whether the fix is on.
  6. Which of our passwords and connected accounts live on that server? Think carrier logins, voicemail-to-email accounts, CRM or scheduling connections, and any keys used by an AI receptionist.
  7. If credentials could have been exposed, when will they be changed? Ask for a date.
  8. Will you notify us in writing if you find any sign of misuse on our account?

Keep missed-call handling and AI answering safe meanwhile

Many offices now route missed or after-hours calls to an automated system: a text-back message, a voicemail transcription that lands in a shared inbox, or an AI receptionist that books appointments. These are useful, and they are also extra places where phone data flows. Here's how to keep them running safely while you wait on vendor answers.

Don't switch off what's working without a backup

If your AI receptionist or missed-call text-back runs through a provider that hasn't confirmed a fix, don't panic and pull the plug at the end of a busy day. Instead, decide on a simple fallback: forward after-hours calls to a staff cell or a known answering service for a few days, and put a short voicemail greeting in place.

Limit what the phone system can reach

An AI receptionist usually needs to see your calendar and maybe your customer list. It rarely needs your full email, billing system or patient records. Ask your vendor to narrow its access to what it actually uses. IBM's research found that 97% of breached organizations that had an AI-related security incident said they lacked proper AI access controls. Keeping access tight costs little and is worth doing.

Watch for tools staff added on their own

Front-desk teams sometimes plug in a free call-transcription app or a personal AI assistant to keep up with call volume. IBM's report found that a high level of shadow AI (unapproved AI tools) added USD 670,000 to the global average breach cost. Ask your team what they use. Bring the useful tools into the open and approve them, and drop the rest.

Add a callback rule for money and data changes

Whatever the state of your phone server, adopt a callback rule: any request to change bank details, release records or send payment gets confirmed by calling back a number you already have on file, never the number that just called. It's free, and it protects you against a hijacked phone line as well as ordinary scams.

What it could cost if this goes wrong

Nobody can tell you the exact price of a phone-system breach for your office. What we can do is give you the industry benchmarks so you can size the risk. IBM's 2025 study, as reported by CyberScoop, put the average U.S. data breach cost at $10.22 million, even as the global average fell 9% to $4.44 million. IBM's own summary notes the global figure was down from USD 4.88 million the year before.

Average data breach cost, IBM 2025 study ($ millions)
02.565.117.6710.2210.22U.S. average4.44Global average

Treat those figures as broad benchmarks rather than a forecast of your own office's bill. Still, the timeline is the part every owner should notice: IBM found organizations took a mean of 241 days to identify and contain a breach. That's months of call data, recordings and passwords in the wrong hands before anyone notices. Patching quickly and checking for signs of intrusion is how you avoid becoming part of that average.

The California angle

For example, a business notifying more than 500 California residents of a single breach must send a sample notice to the Attorney General within 15 calendar days of notifying consumers. Medical, legal and financial offices may have additional rules on top of that. Talk to your attorney about what applies to you. Either way, keep a written record of what you asked your vendors and when they answered.

A worked example

Hypothetical: a three-location physical therapy group in Los Angeles (say, Culver City, Pasadena and Burbank) uses hosted cloud phones, a missed-call text-back tool, and an AI receptionist that books new patients after hours. Here's how the owner might handle this week:

  1. Monday morning, 20 minutes: The office manager lists four vendors: the phone provider, the IT firm that set up the desk phones, the text-back tool, and the AI receptionist company. They email all four the eight questions above.
  2. Monday afternoon: The phone provider replies that its platform doesn't run FreeSWITCH. Filed. The AI receptionist vendor says it does, and that the update to 1.11.3 is scheduled for Wednesday night.
  3. Monday evening: The owner asks the AI vendor whether inbound chat can be switched off on their account in the meantime. The vendor confirms enable-chat is now off on every profile that doesn't need it.
  4. Tuesday: The office manager reviews what the AI receptionist can reach. It had read access to the whole shared inbox, which it doesn't need. Access is cut to the scheduling calendar only.
  5. Thursday: The AI vendor confirms the update in writing, states that it found no signs of misuse, and rotates the API key that connects to the clinic's scheduler.
  6. Friday: The owner adds the callback rule to the front-desk checklist and files all four written answers in a "security" folder.

Total staff time: a few hours across the week. No phones went down, and the owner now knows exactly what runs their calls, which they didn't know on Monday.

Pitfalls to avoid

What we'd do for a Los Angeles office

Los Angeles offices often run more phone tools than they realize. Multi-site practices spread across the basin, bilingual call queues, after-hours answering for clinics and trades, and ad call tracking all add vendors to the chain. If we were sitting with you this week, here's the order we'd work in:

  1. Map every system that touches a call, from the main line to the AI receptionist, and who operates each one.
  2. Trim access so voice tools reach only what they actually use.
  3. Rotate the keys and passwords that live on the phone server or connect it to other tools.
  4. Put missed-call handling on solid ground, so after-hours callers still get a fast, accurate response without opening new holes.

That last step is where good automation helps rather than hurts. When we build AI automation for missed calls and front-desk work, we design it so each piece has the least access it needs, credentials are stored and rotated properly, and there's a manual fallback if any vendor has a bad week. If you're weighing a new AI receptionist or want a second opinion on the one you have, an AI consultation is a good starting point, and our Los Angeles page explains how we work with local offices.

Next step

This week, send the eight questions to every vendor that touches your phones, and file the answers. If you'd rather have someone do the mapping and follow-up for you, and check that your missed-call and AI answering setup is safe while it's at it, book a Free Automation Audit. We'll go through your phone and automation stack with you and hand back a plain-English list of what to fix first.

Sources