‘AI-generated scams becoming sophisticated’ – The Nation Newspaper

Artificial intelligence (AI)-generated scams are no longer clumsy phishing emails replete with typos. Today’s fraud is polished, personalized, and emotionally intelligent–and it’s working as people are falling prey to their antics.

Recent data has shown that investment scams dominate the AI-assisted scam, accounting for roughly between 45 and 47 per cent of total losses. Impersonation scams follow at around between 24 and 28per cent, while job-related scams contribute approximately between 10 and 13per cent across major jurisdictions. Cybersecurity experts and regulators have warned that this reflects a structural shift in fraud tactics, signaling a growing risk driven by AI-generated deception.

According to Israeli cybersecurity company with operations in over 60 countries, Check Point Software Technologies, “AI doesn’t just automate scams; it optimizes them.

The most dangerous scams don’t look suspicious at all. They look helpful, calm, and familiar. ”

Check Point’s experts have identified four major baits AI scammers rely on most. They include polite messages that push people to act fast; personal details that create false trust; and generated voices and videos that pressure people. The fourth is perfectly written messages with no real details.

Check Point research shows that over 60per cent of successful phishing attacks now use neutral or friendly language, rather than fear-based messaging.

“AI models are trained to sound cooperative and calm. That tone creates compliance,” Check Point’s fraud prevention specialist said, urging people never to use links or phone numbers provided in the message. “Pause before acting–legitimate requests can wait,” Check Point advised.

On personal details that create false trust, the company said AI-generated scams increasingly include personal details such as people’s name, employer, school, or recent purchases. Studies show that emails containing personal data are up to four times more likely to be clicked than generic messages.

“Check Point’s latest threat analysis shows that personal details don’t make a message legitimate. Scammers routinely use publicly available information to create a sense of familiarity. What matters isn’t how much a message seems to know about you, but whether it can be independently verified through trusted, official channels.

“Personal details do not confirm legitimacy, treat them as neutral information. Verify requests through official channels you access yourself,” the company said.

Addressing the generated voices and videos that pressure people, the company said AI voice cloning and video deepfakes are now used to impersonate family members, school officials, or executives.

“In 2024, financial institutions reported up to 300per cent increase in deepfake-enabled fraud attempts, particularly involving “emergency” requests for money. A major red flag is when the message discourages calling back, verifying, or involving others.

“How to protect yourself:  Treat urgent audio or video requests as unverified by default

Advertisement

“Confirm identity using a pre-established contact method, not the channel used in the message,” Check Point noted.

Finally on perfectly written messages with no real details, it said AI-written scam messages often sound polished and credible, but they avoid concrete details like transaction IDs, case numbers, or official contact channels that would allow verification.

“AI can make a message sound professional and legitimate, but it can’t provide accountability. There’s no customer service trail, no documented transaction, and no responsible party–and that absence is the clearest warning sign,” Check Point’s security strategist said.

On how to protect oneself, it said absence of concrete details is itself actionable information.

“Be cautious of messages that sound professional but say very little. The biggest mistake people make is trusting realism. In the AI era, realism is cheap. Verification is what matters. The safest habit isn’t skepticism–it’s slowing down and checking independently,” Check Point said.