Welcome to “Ask a security trainer,” the column where the digital security training team at Freedom of the Press Foundation (FPF) answers your burning questions at the intersection of journalism and security. Submit yours here! Let’s jump right into this week’s question.

Dear DST,

I’ve heard that AI is being used to make phishing attacks worse, but what exactly does that mean? And what can I do to protect myself against it?

This almost makes me miss when it was just people impersonating other people in their phishing scams.

Signed,

Impersonators Upon Impersonators

–

Hi Impersonators,

Phishing attacks depend on impersonating a trustworthy actor to fool you into doing something against your interests, like giving them your credentials or downloading malware. It makes sense, then, that phishing attacks would leverage large language models. After all, what is an LLM, if not a tool for helping to impersonate someone else’s language?

To that end, we are seeing mounting signals that AI is being used in a significant proportion of phishing attacks, and can also make them more effective.

It’s important to note that phishing doesn’t just happen in your email inbox. Attackers can deliver phishing lures over voice and video conversations, through text messages, QR codes, and web applications, and more. But for starters, let’s talk about email, and how — in spite of the changes happening around us — the core defenses against email phishing still apply.

For example, between October 2025 and March 2026, researchers at KnowBe4 estimate a roughly 86% increase in phishing attacks that leveraged AI in some way. Verizon’s 2026 Data Breach Investigations Report, which examines security incidents detected from November 2024 through October 2025, suggests attackers used generative AI in about 44% of the detected incidents to help develop phishing lures as a point of initial access into a system.

While other methods are also a common way someone will use AI to get initial access (32%), phishing remains on top, Verizon reports. According to a 2025 Microsoft threat intelligence report, “AI-automated phishing emails achieved 54% click-through rates compared to 12% for standard attempts — a 4.5x increase.”

This research points to a variety of automated techniques, including conducting reconnaissance on targets, making more personalized lures, and enabling subtle changes between similar lures (e.g., in email subject lines) to avoid detection. Attackers can even draft multilingual lures. Attackers are also increasingly using these tools to develop vulnerabilities, and to generate iterations on malware payloads that similarly help bypass detection.

Taken together, these techniques make it easier to write plausible-looking emails that appear tailored for you, with information relevant to your organization, for example by impersonating colleagues or referencing real public-facing events where you or your colleagues may be present. Further, because of the scale of automation behind these phishing attempts, attackers can iterate on techniques quickly.

While some email providers like Gmail have the resources to better identify patterns in how attackers are using AI to modify attacks, the defenses you should personally employ haven’t changed.

Look out for these red flags

  • Unusual-looking email addresses. Sometimes phishing email addresses are clearly not related to the entity they are impersonating. But sometimes they may attempt to impersonate an entity using similar-looking email addresses, so look closely. For example, a Dropbox email should come from an “@dropbox.com” email address.
  • URLs that have no connection to the service. Similarly, attackers may attempt to fool you into clicking on a link to a website they control. This may be an obvious fraud, but sometimes they are very clever and realistic. You can always bypass this risk by simply navigating to the website yourself. (e.g., there’s no need to click that Dropbox link; navigate to dropbox.com independently.)
  • Social pressure or a sense of urgency. Phishing attacks depend on getting you to click on things before you’ve thought it through. If you find yourself feeling your nerves light up from an email, take it as a sign to hit pause and think about it.
  • Unsolicited attachments. Yes, we know journalists’ jobs are to open unsolicited attachments. But treat this as a possible red flag! Rather than opening them directly, if appropriate, consider opening dodgy files in a sandboxed environment like Google Drive, which can show a preview of a document or photo. You can also use Dangerzone to make safe copies of many document formats.

And remember to use two-factor authentication. Requiring a second piece of information beyond your password before someone can log in, such as a code sent to your phone, will make attackers’ lives that much harder. Read our guide to setting up two-factor authentication.

To learn more, read our guide to defending against phishing. I wish you the best in wasting the time and resources of any attackers in your inbox — human or otherwise.

Cheers,

Martin Shelton