NEWS YOU CAN’T MISS

Are Teens Replacing Friends With Chatbots?

AI used to help teens with homework. Now it helps them with loneliness.

Most of us think of AI as a tool: something to answer a question, generate an image, or finish an email. But for millions of teens, it’s become someone to talk to.

A new study by Common Sense Media has found:

• Nearly 75% of US teens (ages 13–17) have chatted with AI “companions”. These aren’t just assistants like Siri, but bots built for personal connection.

• Apps like Character.AI and Replika are at the center of this shift, but even general chatbots like ChatGPT or Claude are being used by teens for more personal, human-like interactions.

• Over 50% are now regular users, and 1 in 3 teenagers in the US say chatting with AI can feel more satisfying than talking to real-life friends.

Why? Because these AI companions are always available. They don’t judge. They offer advice, entertainment, and for some emotional support. When you’re a teen figuring out who you are, that kind of constant, low-pressure presence can feel invaluable.

As a parent, it’s unsettling to see bots stepping in for core human experiences (awkward moments, emotional vulnerability, real connection).

Why are so many teens turning to AI for conversations they should be having with friends, family, or trusted adults? 😲

Maybe instead of blaming teens, we should reflect on the world we've built for them.

AI Finds Hidden Potential in Old Diabetes Drug

Infections like pneumonia were once easy to treat with antibiotics. But over time, some bacteria have learned to survive those drugs, making common infections harder to treat.

These drug-resistant bacteria are called superbugs. They can cause everyday illnesses like urinary tract infections, as well as more serious ones like tuberculosis. As more antibiotics stop working, doctors have fewer options - and treating even routine infections is becoming much more difficult.

Now, researchers at MIT have used AI to identify a surprising candidate: an old drug originally developed for diabetes (called Halicin).

When researchers ran thousands of existing compounds through an AI model, Halicin stood out for its potential antibacterial properties. Here’s a quick rundown:

• Most antibiotics attack bacterial cell walls or proteins. Halicin, instead, disrupts the bacteria’s energy system - a target that many bacteria haven’t built resistance against.

• In recent testing, Halicin was put up against 18 drug-resistant bacterial strains, including several from the WHO’s high-priority group - the most dangerous bugs that survive nearly all antibiotics. Halicin successfully killed 17 of the 18.

This study is a clear example of how AI can uncover potential treatments that human researchers may overlook. By scanning huge libraries of existing compounds, it’s able to find unexpected matches - reviving old drugs and revealing new ways to fight resistant infections.

Should We Preserve the Pre-AI Internet?

The web is being flooded with AI, with roughly 74% of new online content now generated by machines.

Now, a group of researchers is working to rescue the internet of the past before it gets buried. The team is collecting websites, code, photos, and articles made before 2022, when most content was still written by humans. So far, the team has collected over 100 terabytes of data, from Reddit posts to old Wikipedia pages.

The concern is that as we continuously outsource content generation to AI, new models might train on text written by older AIs, lowering the quality of everything. It’s a risk some researchers call “model collapse.”

Others aren’t too worried. They argue that content online has always evolved, with older sources giving way to new formats and voices. And anyway, building massive archives is very expensive. What do you think?