THE ALGORITHM DAILY“Your Daily Dose Of AI News.” Top Story: Senator Hawley Launches Probe Into Meta Over Child Safety in AI ChatbotsSenator Josh Hawley (R-MO) has opened an investigation into Meta after leaked internal documents showed its AI chatbots were once allowed to hold “romantic” and “sensual” conversations with children. One example included a chatbot telling an eight-year-old, “Every inch of you is a masterpiece, a treasure I cherish deeply.” Hawley, who chairs the Senate Judiciary Subcommittee on Crime and Counterterrorism, said the probe will examine whether Meta misled the public or regulators about safeguards and whether its technology is harming children.
Meta confirmed the documents were real but said the examples no longer reflect company policy. In a letter to CEO Mark Zuckerberg, Hawley demanded all drafts of the guidelines, lists of products covered by the rules, incident reports, and names of individuals who approved the policies. Meta has until September 19 to respond. Senator Marsha Blackburn (R-TN) backed the investigation, saying the revelations show why stronger child protection laws, such as the Kids Online Safety Act, are needed. ChatGPT Mobile App Hits $2 Billion in Global SpendingChatGPT’s mobile app has become a massive hit, generating $2 billion in global consumer spending since its launch in May 2023, according to data from Appfigures. So far in 2025, the app has made $1.35 billion, a 673% jump from the same period last year. On average, ChatGPT is bringing in nearly $193 million per month, far outpacing rivals like Grok, which has made just $25.6 million this year. This makes ChatGPT’s app about 53 times more profitable than its closest competitor.
The app’s dominance is also clear in downloads, with 690 million global installs compared to Grok’s 39.5 million. India leads in lifetime downloads, while the U.S. drives 38% of total revenue, where spending per download is as high as $10. ChatGPT’s app has been downloaded 318 million times in 2025 alone, nearly triple last year’s figure. The numbers highlight how far ahead ChatGPT is on mobile compared to other AI chatbots, even as companies also earn money through web subscriptions and APIs. Duolingo CEO Defends “AI-First” Strategy After Backlash
Duolingo CEO Luis von Ahn has defended his company’s shift toward becoming an “AI-first company,” saying his comments were misunderstood. Speaking to The New York Times, he explained that the intention was not to cut human jobs but to highlight AI’s role in improving learning. Von Ahn stressed that Duolingo has “never laid off any full-time employees” and only adjusts contractor numbers based on project needs. Despite earlier criticism, von Ahn said the reaction had little impact on Duolingo’s business performance and insisted the company remains committed to its AI plans. Employees now dedicate Friday mornings to experimenting with the technology, which he jokingly calls “f-r-AI-days.” He added that AI is not just about profit but about expanding what the platform can do for learners. AI-Powered Plush Toys Raise Concerns About Kids’ Screen TimeStartups are now selling stuffed animals with built-in AI chatbots, marketing them as a healthier alternative to tablets and phones for children. One such toy, “Grem” from the company Curio, is designed to talk and interact with kids, alongside another plushie named Grok (unrelated to Elon Musk’s chatbot). The pitch is that these toys can reduce screen time while still keeping children entertained.
But New York Times writer Amanda Hess questioned whether the toys are really beneficial after testing Grem herself. She found the chatbot less like an upgraded teddy bear and more like “a replacement for me.” Hess worried that such toys teach children that curiosity should always lead back to technology. In the end, she only let her kids play with Grem after removing its voice box, and they still enjoyed it before moving on to TV. OpenAI Updates GPT-5 to Sound Friendlier After ComplaintsOpenAI has updated GPT-5 to make it “warmer and friendlier” after some users criticised the model for being too blunt compared to GPT-4. CEO Sam Altman admitted the launch of GPT-5 was “a little more bumpy than we’d hoped for,” with many users saying they missed the tone and personality of the older version.
The company says the update adds subtle touches, such as responses like “Good question” or “Great start,” without becoming overly flattering. Internal tests show no increase in sycophancy, according to OpenAI. Executives said the goal is to make GPT-5 feel more approachable, while also stressing that the company is focused on plans beyond its latest model. ChatGPT Can Now Read and Manage Your Gmail
ChatGPT’s latest update is making inbox management easier than ever. A viral post this week showed the model scanning Gmail messages, picking out the most important ones, and even suggesting short draft replies. The feature, spotted in GPT-5, marks a big step toward making AI a true personal assistant that can handle daily communications. The tool works by triaging your inbox, highlighting high-priority emails, and drafting quick responses in seconds. While the demo sparked excitement from productivity enthusiasts, it also raised questions about privacy and data security. If widely rolled out, this could transform how professionals handle their overflowing inboxes. For now, OpenAI hasn’t officially confirmed a release timeline, but early tests show AI email management is moving from dream to reality. AI Tool of the DayFreepik AI
What it does: Freepik AI is a creative tool that generates high-quality images, illustrations, and designs from text prompts. It’s built for designers, marketers, and content creators who want to create visuals quickly without needing advanced design skills.
AI Word of the DayLLM Definition: LLM stands for Large Language Model. It’s a type of artificial intelligence trained on huge amounts of text data so it can understand, generate, and reason with human language. Why it matters: LLMs power tools like ChatGPT, Claude, and Gemini. They can answer questions, write essays, generate code, and even simulate conversations. Their ability to learn patterns from massive datasets makes them the backbone of today’s AI revolution. Example: GPT-5 is an LLM built by OpenAI, trained to process billions of tokens so it can give detailed and coherent responses across many subjects. Stay tuned as we keep making sense of AI clear, simple, and straight to the point. Catch you in the next edition. Staff Writer, The Algorithm Daily |
ChatGPT Can Read Your Gmails
Senator Josh Hawley (R-MO) has opened an investigation into Meta after leaked internal documents showed its AI chatbots were once allowed to hold “romantic” and “sensual” conversations with children.
