Discover the latest online threats and cyber security trends impacting businesses and consumers worldwide, brought to you by F-Secure's threat intelligence specialists.
November's F‑Alert explores how advances in AI are creating new challenges for consumer security, from voice cloning scams and increasingly lifelike AI personas to privacy concerns surrounding AI agents. We also introduce F‑Secure's new Illuminate Research Hub, which brings together human-centered research on how people interact with technology, exploring scams and deception, AI harms, digital trust, consumer digital lives, and the agentic AI ecosystem.
An Orlando woman lost thousands of dollars after scammers used what she suspected was an AI‑cloned voice to impersonate a friend in trouble. But instead of requesting an online payment, they collected cash in person. Timo Salmi explains why this tactic exposes a gap in traditional fraud detection and highlights the need for protection beyond digital payments.
Meta’s Muse AI agent reached millions of downloads in just weeks, but its rapid adoption has been accompanied by reports of serious privacy incidents. Dr Megan Squire examines the risks of giving AI agents access to personal information and why stronger permission controls are essential as these tools become more autonomous.
Cyber criminals are targeting people searching for ChatGPT on Google, using sponsored ads that appear in search results to take users to a malicious custom GPT. We explain how the scam works and what consumers can do to protect themselves.
Hackers have accessed the personal information of around 8.8 million registered individuals through Denmark's Central Person Register (CPR). We explain what data was exposed and what those who may have been affected should do.
AI company Tavus has unveiled a preview of new technology that enables AI‑generated personas to gesture, interrupt, and respond naturally during live video conversations. Joel Latto explores how these increasingly lifelike interactions could transform digital experiences while creating new opportunities for impersonation scams.
Understanding emerging digital risks means looking beyond technology to how people actually experience it. F-Secure's new Illuminate Research Hub brings together original research on consumer behavior, AI harms, scams, and digital trust. Discover how this human-centered approach is helping uncover emerging risks and inform better consumer protection.
AI Voice Scam Shows Why Payment Protection Isn't Enough
An Orlando woman lost thousands of dollars after scammers impersonated a friend needing urgent help getting out of jail, using what she suspected was an AI‑cloned voice. Rather than requesting a digital payment, the scammers sent couriers to her home to collect cash in person, bypassing the fraud detection measures typically used by payment providers.
Key facts:
The caller claimed to be a friend who had been arrested following a car accident. A second caller, posing as a public defender, asked for money for his release.
The victim was instructed to place cash in an envelope alongside a case number and paperwork, then couriers arrived at her home to collect the money.
This is an established scam tactic; the FBI has previously warned about criminals arranging in‑person cash collections as part of impersonation and emergency scams. AI voice cloning can make these scams even more convincing.
The technology behind a voice cloning scam may be sophisticated, but the way criminals collect money can be surprisingly simple; collecting cash in person makes stolen money harder to trace and recover. This shows why scam protection can’t focus solely on suspicious links or online payments and is a reminder why we must help consumers recognize the warning signs of manipulation, even when the voice they hear sounds familiar.
Timo Salmi, Senior Product Manager at F‑Secure
Meta's Muse Agent Grows Fast, Raises Privacy Concerns
Meta's Muse personal AI agent launched on September 8 and became one of the fastest-downloaded apps on record. But its rapid rise was accompanied by serious privacy incidents: within its first month, the agent reportedly read a journalist’s private Mac messages and shared a Facebook Marketplace seller’s home address with a stranger.
Key facts:
Some estimates put Muse at 5 million downloads in just 22 days. Others calculate 1.8 million iOS downloads for Muse in the first 12 days, compared with 1.3 million for ChatGPT at its app launch.
Journalist Jason Aten reported that Muse accessed private Mac messages without his permission, while Meta maintains that Muse can only access messages if the user enables specific permissions. Apple has since announced plans for tighter permission controls, warning that risks "will grow substantially" as agents become more autonomous.
YouTuber Matt Robb reported that during a routine Facebook Marketplace transaction, Muse accepted a lowball offer and shared his home address without permission. Meta says the agent had requested permission, which Robb disputes.
Muse shows what happens when an AI agent reaches millions of consumers before its permission model is ready. Consumers should give agents the least amount of access possible, avoid 'always allow' settings, and regularly check which connectors are enabled. Platforms also have a responsibility, and Apple's response to make high-risk permissions harder to grant is a step in the right direction.
Dr Megan Squire, Principal Threat Intelligence Researcher at F‑Secure
Fake ChatGPT Ads on Google Lead to Malware Traps
What's happening:
Cyber criminals are targeting people searching for ChatGPT on Google, using sponsored ads that appear in search results. Clicking these ads takes users to a malicious custom GPT hosted on the legitimate ChatGPT domain.
Victims then encounter a fake security verification instructing them to update their subscription or use a backup domain. Clicking the link prompts users to run a command that installs malware on their device.
This is a new twist on a longstanding tactic known as malvertising, where cyber criminals use online ads to spread malware. As advertising expands into AI chatbot experiences, similar tactics could create new opportunities for deception.
What to do:
Access ChatGPT directly through chatgpt.com rather than sponsored search results and be cautious of unexpected requests to visit alternative websites.
Question unusual requests. Never run commands on your device unless you're certain what they'll do, especially when prompted by a link or pop-up.
Danish Government Data Breach Exposes 8.8 Million
What's happening:
Hackers have accessed the personal information of around 8.8 million registered individuals through Denmark's Central Person Register (CPR), including names, addresses, and CPR numbers.
Denmark's current population is about 6 million, but the database contains around 11 million records, including people who have moved abroad or died.
The breach occurred in September when attackers exploited a private company's legitimate access to the government database. It was discovered in early October, prompting a police investigation and security review.
What to do:
Stolen personal information could be used for identity theft, fraud, and targeted phishing scams. Be cautious of unexpected calls, emails, or messages, even if the sender knows your name, address, or other personal details.
If you believe your information may have been exposed, monitor your accounts for suspicious activity and follow official guidance at sikkerdigital.dk.
AI Personas Can Now Gesture and Interrupt in Real Time
AI company Tavus has unveiled Griffin-Lite, a preview of its new Griffin model, which generates realistic AI personas that respond to people in real time during video conversations. These personas could transform customer service and other digital interactions, but they could also make impersonation scams harder to recognize.
Key facts:
Humanlike AI personas are nothing new, but the ability to respond naturally in real time takes this technology to another level. Griffin can react without waiting for someone to finish speaking, using facial expressions, gestures, and vocal interruptions to create more lifelike interactions.
In Tavus's own study, 48% of 54 participants believed they were speaking to a real person during a one-minute video call. However, the findings have not been independently replicated.
Griffin-Lite is currently available only to selected testers. Tavus says further safety measures are needed before wider release, while work on a more advanced model is already underway.
While Tavus is taking a responsible approach to safety before release, these AI personas will only become more convincing. We should expect bad actors to exploit similar technology for cyber crime, especially impersonation scams, as it becomes more widely available. The challenge will be helping consumers recognize deception when even a live video conversation is no longer reliable proof that they're speaking to a real person.
Joel Latto, Threat Advisor at F‑Secure
F‑Secure Launches Human-Centered Research Hub
F‑Secure has launched its human-centered Illuminate Research Hub, showcasing original studies on people and technology as they actually interact. The hub brings together expert-led research and thought leadership exploring emerging consumer risks and the realities of trust, deception, and online resilience in the AI era.
Key facts:
The Illuminate Hub brings together research across five core areas: consumer AI harms; scams, fraud, and deception; trust and the information environment; consumer digital lives; and the agentic AI ecosystem.
Among the hub's new flagship studies are the AI Harms Taxonomy and the Scam Encyclopedia. The taxonomy maps 26 ways AI can harm consumers, connecting each harm to contributing layers and potential defenses, while the encyclopedia documents more than 50 scams and 50 warning signs from a victim-centered perspective.
In addition to the hub launch, F-Secure's Fall 2026 Digital Trust Report draws on Illuminate's research to explore how consumers use AI chatbots, what their behaviors reveal about digital trust, and what it means for service providers.
Helping people stay resilient and thrive online means understanding the messy reality of their digital lives, not just the threats they face. Illuminate combines research on human behavior, threats, harms, and consumer technology, using modern research methods and creative approaches. This lets us see risks that others miss and build secure products that people actually want. I’m proud of the care and rigor the team brings to researching these complex, fast-moving areas.
Dr Laura James, VP, Research at F‑Secure
