AI security risks: what hotels need to know and do

Hotels are adding AI to reservations, guest messaging and back-office work, often faster than anyone checks what those tools do with data. That raises a practical question for owners, general managers and IT leads: what are the real AI security risks, and where should a hotel start? The short answer is that AI protects and attacks at the same time, and the weakest point is still people.

Dr. Ram Kumar, a cyber security and risk leader who began his career in information security in January 2004, calls AI both a boon and a curse for the security world. In his view, AI can be a tool, a medium and a target for cyber attacks. This article follows his framework and what it means for any hotel that stores guest data.

Why AI works as both shield and weapon

On the defensive side, AI helps teams spot trouble faster. Dr. Kumar lists speedier threat detection, network protection, anti-phishing measures and behavioral analysis of patterns to defend against common cybercrimes.

“One thing we need to understand as far as security is concerned, AI has been a double edged sword.”

Dr. Ram Kumar, 5:25

The same technology serves attackers. According to Dr. Kumar, AI makes it easy to run phishing campaigns because it drafts flawless emails that are not easy to detect. He also points to malware that keeps mutating and to AI being used to exploit software vulnerabilities.

He adds one distinction worth keeping in mind: security of AI and AI for security are two different things. Using AI to defend your systems does not mean the AI tools themselves are secure. A related article looks at AI travel planning for hotels.

The three layers of AI security risks

Dr. Kumar groups AI security risks into three levels, from how staff use AI tools down to the models underneath them. The list below follows his threat map.

• Usage: sensitive information can leak when people feed it into AI tools, jailbreaking pushes tools past their limits, and shadow AI apps are used without any security or privacy review.
• Application: poorly secured AI agents open the door to indirect prompt injection attacks, data leakage and exfiltration, and insecure plug-in design.
• Platform: the foundation model and its training data can be targeted with data poisoning, and the model itself can be stolen.

For hotels, the first layer is the most urgent. Many AI apps are freely available, and staff who adopt them on their own may share sensitive information with services nobody has evaluated. Dr. Kumar's point is direct: without proper evaluation from security and privacy aspects, free use of these apps leads to threats and risks.

The third layer matters even if a hotel never trains its own model. If the data behind a model lacks integrity, he warns, the results can be disastrous. A related article looks at hotel business intelligence and data management. He also ties security closely to ethics and responsible AI, recalling a case in which a popular AI platform produced an offensive image that caused public outrage.

Why people remain the weakest link

Asked about the most underrated part of cybersecurity, Dr. Kumar answers without hesitation: the human aspect. People are considered the weakest link in the chain because social engineers manipulate them, and in his view AI makes that manipulation easier still.

Social engineering no longer arrives only as an ordinary phishing email. He names spear phishing, voice phishing, voice cloning, deepfakes and business email compromise, and notes that people have lost serious money through deepfakes.

“All of these are fed by human gullibility, human greed, or it could also exploit our good nature, willingness to help.”

Dr. Ram Kumar, 13:10

That last point deserves attention in hospitality, where good nature and a willingness to help are qualities teams are hired for. Curiosity is another trigger: one click on a link can download malware or ransomware. Dr. Kumar warns that no amount of training removes the chance of someone being careless once, and a single incident is enough to bring down the whole house.

Why hotels are an attractive target

In Dr. Kumar's assessment, the hospitality sector has traditionally not been very cybersecurity savvy. He refers to a major global hotel chain, which he chose not to name, that was targeted repeatedly. Hotels hold sensitive guest information, including personal details and payment data, which makes breaching their databases lucrative for hackers.

“Nobody goes to a hotel expecting that their data would be stolen or exfiltrated.”

Dr. Ram Kumar, 27:35

Guests do business with a hotel in strict confidence. When their data ends up on the dark web, including credit card details, he calls it very unsettling. His conclusion is that the hospitality sector should level up and be on par with other industries when it comes to protecting data, privacy and customer information. A related article looks at hotel security and guest trust.

The risks described above sharpen that call. A hotel that lets staff use unvetted AI tools, or that connects AI agents to its systems without securing them, adds new routes to the same valuable data.

How AI security risks will evolve

Looking ahead, Dr. Kumar expects AI-powered attacks to keep appearing in new ways and to grow more sophisticated, while AI-powered security solutions develop alongside them. He highlights several trends worth factoring into any hotel's planning:

• Supply chain risks are getting worse.
• Social engineering supported by AI is becoming very difficult to detect.
• Zero trust is set to be adopted in a big way.
• Cloud risks are increasing, and cloud outages are a worrying factor that could push some companies back to on-premises models.
• Security and privacy regulations keep getting more stringent.
• Hybrid work keeps the security risks of remote working on the agenda.
• The shortage of cybersecurity talent will continue.

He also expects AI to change hospitality jobs directly. Repetitive roles such as receptionists, customer service representatives and salespeople could be heavily affected. His advice is not to panic but to upskill and learn to use AI, because people who keep away from it could be in trouble.

What any hotel can take from this

Taken together, these points translate into a short checklist for hotels of any size:

• Know which AI tools staff use, and evaluate them for security and privacy before they touch guest data.
• Treat AI applications and their plug-ins as systems to secure, not only as tools that add security.
• Prepare teams for convincing phishing, voice cloning and deepfake requests, since AI removes the usual warning signs.
• Bring the protection of guest data up to the level of other industries.
• Help staff learn to use AI instead of avoiding it.

You can watch the full conversation for his complete threat map, his view on AI and jobs, and his advice for people starting a career in cybersecurity.

Frequently asked questions

What are the biggest AI security risks for hotels?

In Dr. Ram Kumar's framework, the risks sit on three levels: how staff use AI tools, how AI applications are secured, and the integrity of the models and training data underneath. For hotels, the most urgent are sensitive guest information leaking into unvetted shadow AI apps and AI-written phishing that is hard to detect.

What is shadow AI?

Shadow AI refers to AI applications that people use freely without any evaluation from a security and privacy perspective. Because so many AI apps are easily available, staff can adopt them on their own and share sensitive information with them. Dr. Kumar names it as one of the main usage risks of AI.

How does AI make phishing more dangerous?

According to Dr. Kumar, AI can draft flawless emails, so phishing campaigns are easier to run and harder to detect. Social engineering now also includes spear phishing, voice phishing, voice cloning and deepfakes, and in his words people have already lost money big time through deepfakes.

Will AI replace jobs in hospitality?

Dr. Kumar expects AI to replace repetitive and monotonous work that can be automated, and sees roles such as receptionists, customer service representatives and salespeople under threat. His advice is not to panic but to upskill and learn to use AI, since people who keep away from it could be in trouble.