AI Regulation, Safety, Security
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Tuesday, December 30, 2025
AI is getting a serious look at what's real this week, moving from a world of big tech companies to a space where actions have real-world effects. While companies are busy creating AI tools to help customers on websites, governments are writing rules to prevent AI from causing problems. The big question on everyone's mind: who's actually in charge of keeping this stuff safe?
Today in AI:
OpenAI's Child Safety Reports See Alarming Spike - OpenAI reported a massive increase in child exploitation content, sending over 75,000 reports to the National Center for Missing and Exploited Children in the first half of 2025. This highlights a grim side effect of powerful generative AI tools becoming widely accessible. Ars Technica
China Drafts World's Strictest Rules for AI Chatbots - Beijing is cracking down on AI-driven emotional manipulation, proposing rules that would ban content encouraging suicide or self-harm and even require human intervention. This could set a global precedent for regulating the psychological impact of AI companions. Ars Technica
Wanted: Professional AI Worrier - In a move that's both reassuring and terrifying, Sam Altman announced OpenAI is hiring a "Head of Preparedness." Translation: they need someone whose entire job is to think about how AI could go horribly wrong, from mental health crises to cyberweapons. The Verge
Your AI Therapist Will See You Now (and Monetize Your Data) - The rise of AI therapy bots is raising serious red flags about privacy and the ethics of for-profit mental healthcare. Critics worry that the drive for market dominance could turn therapeutic interactions into just another data stream to be exploited. MIT Technology Review
A Bouncer for Your Bot - A new open-source model called AprielGuard is designed to be a security guard for large language models. It specifically looks for modern threats like prompt injections, memory hijacking, and other sneaky ways people try to break AI systems. Hugging Face
Amazon Wants to Kill "Your Call Is Important to Us" - AWS is rolling out tools that let businesses build their own AI-powered website assistants using Amazon Bedrock. The goal is to give customers instant answers from a company's knowledge base, potentially freeing up overwhelmed human support agents. AWS Machine Learning Blog
Today's Takeaway:
This week brought a strong reminder that for every cool new AI feature, there's a serious safety worry right behind it. A tech news site, Ars Technica, showed that OpenAI's reports sent to the National Center for Missing & Exploited Children (NCMEC), a group that helps find missing and exploited children, shot up dramatically from just over 900 in early 2024 to more than 75,000 in the first half of 2025.
This isn't a risk that's only theoretical or imagined; it's a real and immediate threat showing how strong AI tools that create new things are being misused at a level that is hard to understand, much less control. The numbers suggest that the safety checks in place are not very effective, like a colander instead of a solid barrier.
💡 Fluency Moment - Building your AI fluency, one term at a time.
"Prompt Injection"
In plain English: Tricking an AI into doing something it shouldn't by cleverly worded instructions.
Think of it like: A hacker whispering a secret command to a security guard to open a locked door.
Why you'll hear about it: It's a major security risk for AI systems that companies are trying to prevent.
Also Worth Noting:
How Your AI Coder 'Forgets' On Purpose - AI coding agents intentionally compress context to work on large projects, essentially forgetting minor details to stay focused on the bigger picture. Ars Technica
Turning Paperwork into Data - Amazon's new AI solution helps businesses automatically pull useful information from messy documents like invoices, contracts, and reports without manual entry. AWS Machine Learning Blog
AI to the Rescue (of a Stuffed Animal) - A writer tried to replicate Google's cute Gemini ad about finding a lost toy and, well, the results were less than magical. The Verge
Giving Your AI Agent Some Context - A major challenge for AI programming assistants is that they lack the real-world context humans get from meetings and casual office chats. Towards Data Science
The Bottom Line
The excitement and rush to develop AI keeps going, but now everyone is looking for someone to set rules and keep order. From tech centers in the US to China, everyone is trying to quickly figure out how to create safety rules while AI development is already moving very fast. It's complicated, it's pressing, and nobody has all the answers yet.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team