AI Scams Are Getting Smarter, and Healthcare AI Has a Blind Spot

We’re in a new era of AI-driven scams

When ChatGPT dropped in late 2022, it didn’t take long for cybercriminals to realize what they had. Suddenly, anyone could generate convincing phishing emails without the telltale grammar mistakes that used to tip off victims. Since then, the playbook has expanded fast: turbocharged phishing, hyperrealistic deepfakes, automated vulnerability scans. AI makes all of it faster, cheaper, and easier to scale.

Organizations are already struggling to keep up with the sheer volume of attacks. And this is only going to get worse as more criminals adopt these tools and the underlying models improve. Rhiannon Williams has the full story on how AI is reshaping cybercrime, and it’s one of the 10 Things That Matter in AI Right Now at MIT Technology Review. Subscribers can also watch a roundtable discussion with AI reporter Grace Huckins and editors Amy Nordrum and Niall Firth breaking down what’s on the list.

Meanwhile, on the other side of the AI coin, we have healthcare. Doctors are using AI for notetaking, combing through patient records to flag people who might need specific treatments, and interpreting X-rays and lab results. A growing pile of studies suggests many of these tools can deliver accurate results. But here’s the uncomfortable question nobody seems eager to answer: Does any of this actually translate into better health outcomes for patients?

Jessica Hamzelou digs into that problem in The Checkup newsletter, and the answer is basically: we don’t know yet. There’s a gap between a tool performing well in a controlled study and actually making a difference in a real hospital setting. It’s the kind of blind spot that should make us pause before rolling out these systems too broadly.

Also worth noting: DeepSeek just unveiled its long-awaited new AI model, DeepSeek-V4, in preview form. The Chinese company says it’s the most powerful open-source platform yet and claims it rivals top closed-source models from OpenAI and DeepMind. It’s also adapted for Huawei chip technology, which is a telling move given the ongoing chip restrictions. CNN, Bloomberg, and SCMP all have coverage, and it’s worth watching how this plays out in the open-source vs. closed-source debate.

On the social media front, more countries are moving to restrict children’s access. Norway is set to enforce the latest ban, the Philippines could follow soon, and there’s growing pushback in the US to get AI out of schools. These are messy, complicated policy debates, but the trend is clear: governments are starting to act, even if the approaches vary wildly.

Two big themes running through all of this: AI is getting more powerful and more accessible, but we’re still figuring out how to handle the consequences. Whether it’s criminals weaponizing the tech or doctors deploying it without solid evidence of benefit, the pattern is the same. The tools are here. The understanding is lagging.