Lightnews — Scholar-powered news

Reposted by Megan Morrone

Kashmir Hill @kashhill.bsky.social · 22d

I'm on The Daily today talking about my reporting on AI chatbots and how heavy users of ChatGPT can go into delusional, and sometimes tragic, spirals: www.nytimes.com/2025/09/16/p...

Also today, an OpenAI blog post from Sam Altman on "safety, freedom & privacy": openai.com/index/teen-s...

Trapped in a ChatGPT Spiral

www.nytimes.com

9 28

Megan Morrone @megmorrone.bsky.social · Sep 2

OpenAI is rolling out new ChatGPT guardrails for teens and people in distress: parental controls, crisis support, trusted contacts.

This is nearly one year after the first parent sued an AI company over a child’s suicide.

Read @axios.com → axios.com/2025/09/02/chatgpt-openai-mental-health-teens

OpenAI outlines new mental health guardrails for ChatGPT

The company says parents will soon be able to link their accounts to those of their teens.

axios.com

Megan Morrone @megmorrone.bsky.social · Aug 21

That’s [email protected]

Megan Morrone @megmorrone.bsky.social · Aug 21

I want to talk to you about your kids’ K-12 new school AI policies this year. For an @axios.com story. Please share this most with anyone. I’m at Megan.Morrone@Axios

1 2

Megan Morrone @megmorrone.bsky.social · Aug 7

I'm looking to chat with anyone who definitely would not like their next of kin to reanimate them with AI after they die...as one does.

This is for an Axios story. DM me here or [email protected].

Megan Morrone @megmorrone.bsky.social · Jul 21

DM me about your AI job anxiety. I'm less empathetic than ChatGPT and it might be used in a story with your permission. So email if you're OK with that: [email protected]

1 3

Reposted by Megan Morrone

Axios @axios.com · Jul 21

NEW data first reported by Axios:

• ChatGPT users send more than 2.5B prompts globally each day, per OpenAI
• More than 330M of those are in US
• The free version of ChatGPT is used by the vast majority of the platform's 500m weekly active users

Altman plans D.C. push to "democratize" AI economic benefits

Altman wants to continue driving debate over the effect AI will have on jobs.

www.axios.com

7 6 21

Megan Morrone @megmorrone.bsky.social · Jul 17

Agents. Agents. Agents. Agents.

ChatGPT Agent is out today. OpenAI says it can start and finish complex web and computing tasks without you holding its hand.

I’ll be testing it soon. In the meantime, here’s my Axios piece with the details:
www.axios.com/2025/07/17/c...

New ChatGPT Agent promises to do more for you

OpenAI says its new tool can take over more of your tasks from start to finish.

www.axios.com

Reposted by Megan Morrone

Sam Sabin @samsabin.bsky.social · Jul 15

CISA spent years building out its role (and brand!) as the country's go-to partner on cyber issues.

piece by piece, that work is being dismantled as workforce reductions and budget cuts hit the agency.

more from today's @axios future of cyber newsletter:
www.axios.com/2025/07/15/c...

The top U.S. cybersecurity agency is shrinking quickly under Trump 2.0

Companies and critical infrastructure will be more vulnerable as a result, experts warn.

www.axios.com

3 4

Megan Morrone @megmorrone.bsky.social · Jul 11

ChatGPT’s new memory feature is here. And it raises big questions.

For @axios.com I talked with OpenAI's Christina Wadsworth Kaplan and Samir Ahmed about how memory works in ChatGPT and what it means to be remembered by a chatbot.

www.axios.com/2025/07/11/c...

ChatGPT keeps having more memories of you

OpenAI's expanded memory feature presents thorny privacy, personalization and control dilemmas.

www.axios.com

1 2

Megan Morrone @megmorrone.bsky.social · Jul 1

I am a very simple woman. All I want is for AI to do my expense reports

1

Megan Morrone @megmorrone.bsky.social · Jul 1

New @axios.com story: Upwork data shows a rise in “vibe coders” — people using AI to code without knowing how — and demand for “generalists” who can blend AI with creative skills.

www.axios.com/2025/06/30/a...

Exclusive: The rise of the AI-savvy generalist

New data from freelance platform Upwork shows how AI is shifting skills.

www.axios.com

Megan Morrone @megmorrone.bsky.social · Jun 26

Anthropic just dropped new data on how folks are using Claude for emotional support.

Bots can be comforting. But they weren’t built for therapy. Thx to @miles.land &‬ Alex Sanderford from @anthropic.com & @erictopol.bsky.social for insights. In @axios.com AI+.

www.axios.com/2025/06/26/a...

Exclusive: New Anthropic report details How Claude became an emotional support bot

The research shows users grow more positive while chatting with Claude.

www.axios.com

1 4

Reposted by Megan Morrone

Chuck McCutcheon @chuckmccutcheon.bsky.social · Jun 12

From @megmorrone.bsky.social: Trump will attend a summit next month at Carnegie Mellon -- a big win for Sen. Dave McCormick's efforts to bring AI and energy leaders together. www.axios.com/2025/06/12/t...

Scoop: Trump to headline Sen. McCormick's AI energy summit in Pittsburgh

Trump's attendance is a win for McCormick's efforts to bring AI and energy leaders together.

www.axios.com

2 3

Megan Morrone @megmorrone.bsky.social · Jun 12

Here is your friendly reminder that LLMs are simply guessing at the next word. They're not designed to give you facts or to verify them. They don't "think," they infer. As far as I know, dogs still are unable to use Google when they've lost their socks.

3

Megan Morrone @megmorrone.bsky.social · Jun 11

I wrote for @axios.com about the executive “digital twin”: bots trained on a leader’s voice, writing, and decisions to scale their time and influence.

With insight from @levie.bsky.social, @samliang.bsky.social @jlevitt.bsky.social and @jgcarpenter.com.

Read here: www.axios.com/2025/06/10/a...

CEOs are scaling themselves with AI digital twins

Execs are pushing AI, but slow adoption at the top can cause workplace rifts.

www.axios.com

2 5

Megan Morrone @megmorrone.bsky.social · Jun 5

As far as I'm concerned this is the biggest news of the day www.axios.com/local/chicag...

Ferris Bueller's iconic vest hits the auction block

"Bueller? Bueller? Bueller?"

www.axios.com

1 3

Megan Morrone @megmorrone.bsky.social · Jun 5

I’ve been on a plane from NY to SF for the past few hours and I’ve never before wished so hard that I didn’t have wifi.

1

Megan Morrone @megmorrone.bsky.social · Jun 4

AI companies could reduce hallucinations, but it would make their models slower, costlier, and less “helpful.”

So they optimize for speed and scale instead.

That’s why chatbots still confidently make stuff up. And probably always will.

Megan Morrone @megmorrone.bsky.social · Jun 4

Some highlights:

– RAG and SAFE aim to reduce hallucinations, but can’t eliminate them
– Newer models hallucinate more because they reason more
– “Hallucination-free AI may never arrive”

1 1

Megan Morrone @megmorrone.bsky.social · Jun 4

For @axios.com, I wrote about why AI hallucinations persist. And why tech companies aren’t doing more to fix them. (Hint: accuracy is expensive, and speed sells.)

🧵 www.axios.com/2025/06/04/f...

AI's "making stuff up" problem won't go away

Companies are making some headway at limiting "hallucinations." The price is slower, costlier AI.

www.axios.com

2 1

Megan Morrone @megmorrone.bsky.social · May 29

The ChatGPT bans are gone, but the fear and secrecy stuck around.

People are using AI behind their bosses backs, and it’s a recipe for screwups.

New from me at @axios.com , with insight from Gong, Coursera, @atlassian.bsky.social, mixus and @emollick.bsky.social: www.axios.com/2025/05/29/s...

Workers are hiding their use of ChatGPT and other AI

Why employees won't admit to using generative AI.

www.axios.com

5 14

Megan Morrone @megmorrone.bsky.social · May 23

Brace yourselves: We've officially climbed out the other side of the uncanny valley.

There’s something eerie and also inevitable about watching an AI-generated actor wrestle with the fact that they were created by a prompt.

My latest for Axios:
www.axios.com/2025/05/23/g...

Google's new Veo 3 AI video tool floods internet with real-looking clips

Many seem to feel this is the moment we stop being able to tell the difference between real and AI.

www.axios.com

5 8 21

Megan Morrone @megmorrone.bsky.social · Apr 25

I talked to experts about what we lose when we turn to prompts instead of people.

Story here: www.axios.com/2025/04/25/a...

Chatbot mentoring is popular — but relying on it is risky

Workers and students who turn to AI instead of their teachers and bosses can miss out on key lessons.

www.axios.com

1 5

Megan Morrone @megmorrone.bsky.social · Apr 16

OpenAI's newest model "thinks" with images, per @inafried.bsky.social on @axios.com www.axios.com/2025/04/16/o...

OpenAI rolls out o3 and o4-mini

The new models are designed for better reasoning and visual understanding.

www.axios.com

1 6