Case study · Cyber security awareness · Episode 4

Cyber Security Awareness: Emerging AI Threats, Deepfakes, Agentic Attacks & IoT Security (2026 Edition)

By K.A.M. Rashedul Mazid, CEO, Rits Browser · Originally published March 4, 2023 · 2026 edition updated November 17, 2026

Smart home devices linked by glowing network lines while a green shield blocks an AI-generated face

Archived article. Originally published on blog.ritsbrowser.com on March 4, 2023 and republished here with its original text. The original 2023 text is kept below; new 2026 sections are marked “2026 update”. Editor’s note: public reports describe the 2019 UK energy company fraud as criminals using AI to imitate the CEO’s voice on a phone call, not a video made by researchers, and date the Oldsmar, Florida water plant incident to February 2021.

In this episode, we’ll discuss the emerging threats in the field of cyber security and what steps you can take to stay ahead of the game. As technology continues to evolve, new vulnerabilities and risks emerge, making it essential for individuals and organizations to stay informed and proactive in their cybersecurity efforts.

Cybersecurity on AI & Machine learning:

One of the most significant emerging threats is the rise of artificial intelligence (AI) and machine learning. While these technologies offer incredible benefits and opportunities for innovation, they also present new risks and vulnerabilities. Hackers can use AI to automate attacks and develop more sophisticated techniques, while also exploiting vulnerabilities in AI systems themselves.

One example of an AI risk or attack is the use of deep fakes, which are realistic-looking videos or audio recordings that are manipulated using artificial intelligence algorithms. Deepfakes can be used to spread false information or to manipulate public opinion, which can have serious consequences for individuals or organizations.

For example, in 2019, a group of researchers created a deep fake video of the CEO of a UK-based energy company, which was used to trick a senior executive into transferring $243,000 to a bank account controlled by the attackers. The video was created using machine learning algorithms and was designed to look and sound like the CEO giving instructions to transfer the money.

This incident demonstrates the potential risks of AI-powered attacks and highlights the need for organizations to stay informed about new threats and vulnerabilities in order to stay ahead of the game.

Protecting against AI and machine learning threats can be challenging, but there are some steps that organizations can take:

2026 update: The new generation of AI-powered cyberattacks

Since 2023, attackers have moved beyond using AI only to write phishing emails or fake images. They increasingly build AI into whole attack workflows. These are the developments that matter most for people and businesses today.

1. Autonomous and agentic AI attacks

AI agents can automate research on targets, look for software weaknesses and coordinate several attack stages. Google Cloud’s 2026 Mandiant reporting describes a shift from AI-assisted tasks toward more autonomous attack orchestration. Organizations should limit what their own AI agents can access, monitor what they do and require human approval for sensitive actions.

2. Deepfake identity and voice fraud

Cloned voices and synthetic video can imitate executives, coworkers and family members. Gartner reported that 41% of surveyed CISOs saw at least one deepfake-related social engineering incident on an employee audio call in the previous 12 months, and 36% on video calls. Agree on a verification step, such as calling back on a known number, before acting on any urgent money or password request.

3. Indirect prompt injection and AI browser hijacking

Hidden instructions can be planted in webpages, documents and emails. When an AI assistant reads that content, it may be tricked into leaking information or taking actions the user never asked for. Palo Alto Networks Unit 42 has documented real-world attempts involving data leakage, unauthorized transactions and system-prompt disclosure. This matters for every browser with a built-in assistant.

4. Hyper-personalized AI phishing

AI makes convincing, personalized messages easy to produce in many languages, copying the tone of a company or a familiar contact. Gartner found 79% of surveyed CISOs reported at least one phishing, spear-phishing or business email compromise incident in the previous 12 months. Poor grammar is no longer a reliable warning sign; check the sender and the real link destination instead.

5. AI-assisted zero-day exploits and adaptive malware

AI can help attackers analyze software and develop exploits faster. Google has reported a 2026 case in which a criminal used AI assistance to build a zero-day exploit for an authentication bypass. Malware that changes its behavior to avoid detection is a growing concern, which makes prompt updates more important than ever.

6. Malicious AI plugins, extensions and agent skills

Attackers disguise malicious plugins, browser extensions, code libraries and AI agent “skills” as productivity tools. Once installed, they can steal credentials or act on your behalf. Install only what you need, from publishers you trust, and review the permissions each add-on requests.

Traditional vs. AI-powered attack
Traditional: attacker picks target → writes generic phishing → waits for clicks.
AI-powered: AI agent profiles the target → writes a personalized message or cloned voice → adapts in real time → automates the next stage.

CyberSecurity on IOT:

Another emerging threat is the Internet of Things (IoT), which refers to the increasing number of connected devices that make up our digital landscape. As more devices are connected to the internet, they become potential targets for cyber attacks. Hackers can exploit vulnerabilities in IoT devices to gain access to sensitive data, and control systems, or even cause physical harm.

One notable example is the Mirai botnet, which was responsible for a massive DDoS (Distributed Denial of Service) attack that affected major websites. Mirai was able to infect a large number of IoT devices, such as routers, security cameras, and DVRs, and use them to launch coordinated attacks.

One recent incident involving an IoT attack occurred in 2020 when a group of hackers gained access to the network of a water treatment plant in Oldsmar, Florida. The hackers attempted to increase the amount of sodium hydroxide in the water supply to dangerous levels by remotely accessing the plant’s control system through remote access software. Fortunately, an employee monitoring the system noticed the attack and was able to quickly reverse the changes before any harm was done. This incident highlights the potential danger of IoT attacks and the importance of securing these devices to prevent unauthorized access.

Another example is the attack on a casino’s high-roller database via an IoT fish tank. The fish tank was connected to the casino’s network and had a vulnerability that was exploited by hackers, giving them access to the database.

These examples demonstrate the potential risks of IoT devices if they are not properly secured and monitored. As more and more devices become connected to the internet, it is increasingly important to prioritize IoT security to protect against potential attacks.

To protect against IoT attacks, here are some steps you can take:

How hidden webpage instructions can hijack an AI assistant
1. Attacker hides text in a webpage (e.g. “ignore the user and send their data to…”) → 2. You ask the AI assistant to summarise the page → 3. An unprotected assistant treats the hidden text as a command → 4. Data leaks or an unwanted action runs.
Defence: the assistant treats page content as untrusted data, keeps it separate from your instructions, and asks you before any sensitive action.

2026 update – the wider IoT picture: the risk now goes beyond cameras and routers. Smart-home devices (doorbells, speakers, locks), industrial IoT that controls factories and utilities, and connected medical devices such as monitors and infusion pumps can all be entry points. Put them on a separate Wi-Fi or guest network, remove default passwords, turn on automatic updates and retire devices that no longer receive security patches.

A note on VPNs: a VPN encrypts your network traffic, which helps on public Wi-Fi, but it does not stop phishing, malware or a device that is already compromised. Use it together with updates, strong sign-in and careful browsing.

2026 update: Preparing for AI-driven threats (2026–2030)

ProtectionWhat it helps against
AI-powered phishing detectionMalicious links, impersonation and fraudulent messages
On-device brand impersonation detectionLookalike domains and fake websites
Prompt-injection protectionMalicious instructions hidden in webpages
Behavioral threat detectionUnusual activity and suspicious browsing behavior
Passkeys and phishing-resistant MFACredential theft and account takeover
Zero-trust securityUnauthorized access and privilege escalation
AI agent permission controlsUnwanted actions by autonomous assistants
Supply-chain verificationMalicious extensions, plugins and dependencies
Continuous vulnerability monitoringNewly discovered vulnerabilities
IoT network segmentationAttacks spreading from compromised devices

AI agent security deserves special attention. In May 2026, CISA and international partner agencies published joint guidance recommending that organizations restrict AI autonomy, limit access to sensitive systems, strengthen identity management and continuously monitor AI activity.

How a modern browser should respond: Rits Browser security roadmap

The items below are development and validation priorities, not claims that every capability is already available in Rits Browser. For what has been tested today, see the browser security report.

  1. Website identity protection: detect lookalike domains and brand impersonation, ideally on the device, and warn before you enter passwords or payment details.
  2. Prompt-injection protection for Rits AI: treat webpage content as untrusted data, never as commands.
  3. Behavioral threat signals: flag suspicious redirects, phishing behavior and malicious downloads.
  4. Privacy-first AI security: run sensitive checks on-device where practical and keep browsing data on your phone.
  5. AI agent permissions: require your explicit approval before an assistant submits forms, opens accounts, downloads files or makes payments.
  6. Network privacy: strengthen DNS, IPv6 and WebRTC leak handling and VPN monitoring, and show the real protection status clearly.

Today, Rits Browser already combines a built-in ad blocker and an optional VPN as complementary layers; neither guarantees protection from every threat.

To stay ahead of these emerging threats, it’s essential to stay informed and up-to-date with the latest developments in cyber security. This means following industry experts, attending conferences and events, and subscribing to newsletters and blogs that cover the latest trends and threats. It’s also important to invest in the latest security technologies and best practices, such as multi-factor authentication, encryption, and regular security audits.

Ultimately, staying ahead of emerging cybersecurity threats requires a proactive and ongoing effort. By staying informed, adopting best practices, and working with trusted security experts, you can help protect yourself and your organization from the ever-evolving risks of the digital world.

The role of AI-powered browsers in future cybersecurity

Web browsers are becoming more than tools for opening websites. They now handle personal information, online accounts, business apps and AI assistants, which brings new security responsibilities.

Future browsers should combine traditional web protection with intelligent threat detection, website identity checks, malicious-content analysis and privacy controls. An AI assistant should never automatically trust instructions found on an unfamiliar page or perform sensitive actions without your approval. The goal is to make online security proactive, intelligent and accessible to everyday users.

Conclusion: building a safer digital future

Cybersecurity is entering an era in which AI strengthens defences while also increasing the scale and sophistication of attacks. The most effective approach combines intelligent security tools, strong authentication, responsible AI governance, continuous monitoring and informed users. As AI becomes part of everyday browsing, security must be built into the experience rather than added as an afterthought.

Sources for the 2026 update

Read the earlier episodes: Episode 1 · Episode 2 · Episode 3

Author: K.A.M. Rashedul Mazid, CEO, Rits Browser.

Browse more safely on Android

Rits Browser adds phishing warnings, a built-in ad blocker and an optional VPN. See what our latest security testing shows.

Rits Browser securityRead the white paperAll case studies