Keeping situational awareness and response running when the systems behind them fail.

October is Cybersecurity Awareness Month, and most messaging will focus on phishing, passwords, and data breaches. Right now, the loudest conversation is about agentic AI: the tools that act across systems on their own. Not much is being said about what it means for a SOC.

For those who lead security operations centers (SOCs) and the leaders of public safety departments in higher education that deploy them, there is an urgent conversation to be had about the security of the systems that provide situational awareness in the public spaces, research areas, and crowded venues where safety and rapid response to emergencies are balanced with privacy and civil liberties. Regardless of whether the operational model is a SOC or a real-time crime center (RTCC), the risk is the same. Every camera, card reader, video server, and AI tool feeding that center is a potentially vulnerable networked device, and cyber attackers are aware of that.

Recent incidents that apply to a SOC.

While none of the incidents here was a sophisticated attack on a college or university SOC, each highlights a vulnerability for which many higher education safety and security programs have not fully accounted:

Cameras. In February 2026, CISA warned of a critical flaw, rated 9.8 out of 10, in several Honeywell CCTV models that could let an attacker with no password take over accounts and view live feeds. No exploitation had been reported at the time, which was not reassuring, because once flaws go public, attackers look for devices nobody has patched or isolated. Cameras and card readers are easy targets because they are just the kind of devices that get installed and forgotten.

Alerting. In November 2025, a ransomware attack damaged the CodeRED emergency notification platform used by law enforcement agencies and municipalities nationwide. Dozens of agencies went about two weeks without it, and attackers leaked users’ passwords as well as personal details. Many campuses rely on comparable platforms for emergency alerts.

Dispatch. As the fall 2025 semester opened, more than two dozen campuses received hoax active shooter calls. Swatting calls typically rely on temporary phone numbers and voice-cloaking apps, and some of these included background noise that mimicked gunfire. The fastest way to separate a hoax from a real threat is often live video, which makes camera integrity a dispatch, officer safety, and public safety issue, not just an IT issue.

Vendors. In spring 2026, the extortion group ShinyHunters breached Instructure’s Canvas platform, used by more than 40 percent of North American colleges and universities. The attackers claimed data from nearly 9,000 institutions, a figure not independently confirmed. Canvas is not a security system, but the pattern applies to a SOC: one vendor compromise reaches thousands of customers at once. Video management, access control, computer-aided dispatch (CAD), and mass notification platforms carry the same concentration risk.

AI. In October 2025, an AI gun detection system at a Maryland high school flagged a student’s bag of chips as a possible firearm. The district’s security team reviewed and canceled the alert, but the principal had not seen the cancellation and escalated to police, who handcuffed and searched the student. The technology sent the alert to a human as designed, but the handoff failed.

AI in the SOC: force multiplier and new attack path.

AI is arriving in campus SOCs: video analytics that flag weapons or loitering, tools that triage alarms, and increasingly “agentic” assistants that take actions in partnership with an operator, such as pulling footage, drafting reports, or querying systems. Used well, these tools extend the capabilities of a small team. They also add two kinds of risk.

• Operational: false alerts and failed handoffs, which is why every AI alert needs a human who makes the key decisions, with a defined path to confirm an alert and a defined path to cancel it.
• Cybersecurity: In August 2025, researchers from Tel Aviv University, Technion, and SafeBreach showed that a single poisoned calendar invite could hijack Google’s Gemini assistant and make it open window shutters, switch off lights, and turn on a boiler in a smart apartment. The hidden instructions triggered when the user simply asked for a summary of the week. Google says it has fixed those flaws, but the lesson carries directly into the SOC: an AI agent that can pull footage, unlock doors, or query dispatch data could be turned by content it reads, such as an email, a calendar entry, or a work order.

That makes AI governance a core cyber hygiene issue for the SOC and RTCC, on par with patching and passwords. Before any agentic tool touches operational systems, the SOC needs a written policy on what the tool may access, what it may do without a human’s approval, who owns, monitors, and is ultimately accountable for it, how its actions are logged, and how it gets shut off during an incident. SOC leaders should treat an AI agent like a new operator: a defined role, limited permissions, close supervision, and a place in the incident response plan.

Don’t wait for an invitation to the cybersecurity conversation.

In the higher education arena, SOC staff can be the first to encounter cyber incidents yet are too often the last to be invited into the rooms where they are managed. SOC Directors and public safety leaders cannot afford to wait for an introduction or invitation to the cybersecurity conversation. They should show up at the CISO’s door, even if only virtually, with a device inventory, a list of the systems their operators cannot work without, and a request for a joint plan. The cyber team knows networks while public safety and security knows what happens during emergent situations when situational awareness is lacking. The joint playbooks that come from even basic of conversations on cybersecurity are invaluable in mitigating the impact of cyber incidents and ensuring continuity of essential operations during the recovery process.

Train, plan, and exercise together.

Most SOC operators work inside software platforms, including video management, access control, and dispatch systems, and are trained to act on what those platforms show. Too few are trained to recognize when the systems themselves are the incident. Operators should be able to spot the warning signs: feeds that freeze or loop, doors that stop responding, alarms that flood or go silent, and AI alerts that don’t match the screen. They should know exactly who to call, and what not to touch, so response is quick and both systems and evidence are preserved.

That knowledge belongs in a joint playbook written with cybersecurity professionals, covering three phases:

• Identifying a potential cyber incident. How the SOC reports suspicious behavior, how cyber confirms it, and who decides whether to take a system offline.
• Fallback and continuity. Which doors fail locked or unlocked, how patrols, radios, and manual lockdown procedures take over, and when the EOC activates. The test: can you keep campus safe for 72 hours with degraded or untrusted systems?
• Restoration. How systems are cleaned, verified, and reconnected, and how the SOC and the cyber team work together to confirm they can be trusted again.

The bottom line.

Physical security and cybersecurity can no longer plan in separate rooms. The campuses that handle the next incident well will be the ones whose public safety and security leaders walked into the cyber conversation first and decided, both in advance and together, how the SOC keeps working when its own tools are the target.

If you want to talk through how these questions apply to your campus, Sentinel welcomes the conversation.

Sources

CISA, ICS Advisory ICSA-26-048-04, Honeywell CCTV Products (February 17, 2026): https://www.cisa.gov/news-events/ics-advisories/icsa-26-048-04. Secondary: Security Affairs (February 19, 2026): https://securityaffairs.com/188234/security/cisa-alerts-to-critical-auth-bypass-cve-2026-1670-in-honeywell-cctvs.html

Rob Wright, Dark Reading, "CodeRED Emergency Alert Platform Shut Down Following Cyberattack": https://www.darkreading.com/cyberattacks-data-breaches/codered-emergency-alert-platform-shut-down-cyberattack

CNN, campus hoax shooter calls (September 21, 2025): https://www.cnn.com/2025/09/21/us/university-swatting-incident-college-calls

NBC News, Colleges across the country deal with shooting hoaxes as classes resume (August 2025): https://www.nbcnews.com/news/us-news/colleges-shooting-hoaxes-rcna227136

Reed Smith, Canvas/Instructure cyberattack: key developments for higher education (June 18, 2026): https://www.reedsmith.com/articles/canvasinstructure-cyberattack-key-developments-and-action-items-for-higher-education-institutions/

Karina Tsui and Zoe Sottile, CNN, AI security system mistakes chip bag for gun (October 2025), syndicated: https://abc11.com/post/student-handcuffed-doritos-bag-mistaken-gun-schools-ai-security-system-baltimore-county-maryland/18073796/

Google, AI threats in the wild: The current state of prompt injections on the web (April 2026): https://blog.google/security/prompt-injections-web/. Independent coverage: SecurityWeek (April 2026): https://securityweek.com/malicious-ai-prompt-injection-attacks-increasing-but-sophistication-still-low-google

Dark Reading, Google Gemini AI Bot Hijacks Smart Homes, Turns Off the Lights (August 2025): https://www.darkreading.com/cyberattacks-data-breaches/google-gemini-ai-bot-hijacks-smart-homes

CISA, NSA, FBI, and international partners, Principles for the Secure Integration of Artificial Intelligence in Operational Technology (December 3, 2025): https://www.cisa.gov/resources-tools/resources/principles-secure-integration-artificial-intelligence-operational-technology