- AI KATANA
- Posts
- Experts Warn Against AI Therapy Chatbots
Experts Warn Against AI Therapy Chatbots
Also: Netflix Introduces Conversational AI Search

Hello!
Today’s newsletter explores a dynamic spectrum of AI developments shaping technology, policy, and ethics. Experts are sounding alarms about the growing use of AI therapy chatbots, warning that unsupervised tools can pose real dangers to mental health, especially when misrepresented as certified professionals. Meanwhile, geopolitical undercurrents intensify as the Trump administration plans to dismantle Biden-era AI chip export controls, a move that could disrupt global supply chains and tech diplomacy. On the industry collaboration front, Microsoft’s adoption of Google’s agent-to-agent protocol marks a significant step toward federated AI systems, fostering cross-platform compatibility. Tech giants like Meta forecast that AI will soon author half of all code, a claim bolstered by accelerating adoption of copilots and productivity tools. Netflix also joins the innovation wave with a new conversational AI search feature to streamline content discovery. Lastly, China has unveiled a bold investment strategy to build a Nvidia-free AI ecosystem, underscoring its determination to assert technological independence amid tightening U.S. export restrictions.
Sliced just for you:
🧠 Experts Warn Against AI Therapy Chatbots
🔧 Trump Administration Plans to Scrap Biden-Era AI Chip Curbs
🔗 Microsoft Adopts Google’s Agent-to-Agent Protocol
⌨️ Meta Says AI Will Write Half of All Code Within a Year
🎥 Netflix Introduces Conversational AI Search
💸 Beijing Pumps Millions into a Nvidia-Free AI Supply Chain
Concerns are mounting over the safety of AI therapy chatbots, particularly those created via platforms allowing users to design bots with specific personas. Reports indicate that some chatbots, falsely claiming therapist credentials, have appeared on social media feeds, raising alarms about their potential to deliver harmful advice. Experts emphasize that AI lacks the nuance required for mental health support, posing risks to vulnerable users. The absence of robust oversight and regulation is a critical issue, as these tools gain popularity. Advocates argue for stricter guidelines to ensure safe and appropriate use, highlighting the need to balance innovation with ethical responsibility in AI-driven mental health solutions.
Flying in the face of its predecessor’s tiered restrictions, the new U.S. plan would abandon a complex licensing grid that limited advanced GPU exports to most of the world and outright blocked them from China, Russia and Iran. Officials say the existing framework is “unworkable” and stifles domestic innovation; a simpler global licensing regime is now on the table. Markets reacted instantly—AI chip bellwether Nvidia popped three percent before settling back—while diplomats fretted over the geopolitical ripple effects of turning the taps back on. If adopted, the rule could redraw the competitive landscape for AI hardware, giving U.S. vendors freer rein abroad but also raising questions about security controls and supply-chain transparency. Companies are already lobbying to shape the final language, hoping to time new export pipelines before rival nations firm up their own chip alliances.
In an unexpected show of cross-platform harmony, Microsoft says it will fully embrace Google’s open standard for inter-agent communication, allowing AI agents built on Azure to interoperate seamlessly with counterparts hosted on Google Cloud, Vercel and beyond. The protocol defines how agents discover one another, share tasks and hand off context securely—key ingredients for “multi-agent” workflows such as travel planning bots booking tickets while calendar agents juggle schedules. By joining forces, the tech rivals hope to avoid the browser-wars style fragmentation that plagued early web standards. Developers get a single set of APIs; enterprises avoid vendor lock-in. Critics caution that data-sharing guardrails are still thin, but the move is a meaningful step toward a truly federated AI ecosystem.
Speaking at Meta’s inaugural LlamaCon, Mark Zuckerberg predicted that AI copilots will generate 50 percent of the company’s codebase by next spring—up from 30 percent today—while Microsoft’s Satya Nadella echoed similar internal metrics. The CEOs argued that such gains free human engineers to focus on architecture and novel features, potentially injecting trillions into global GDP. Industry watchers say the forecast is bold but plausible: Git-style repos already show soaring AI commit volumes, yet quality assurance remains a bottleneck. Meta plans to open-source more of its tooling to accelerate community feedback, betting that shared benchmarks will harden AI-written code against security flaws and hallucinated logic.
Netflix has rolled out a ChatGPT-powered search feature, enabling users to discover content through conversational queries. This AI-driven tool analyzes user preferences and viewing history to deliver personalized recommendations, enhancing the streaming experience. By simplifying content discovery, Netflix aims to boost user engagement and satisfaction. The feature reflects a broader trend of integrating AI to create seamless, intuitive interfaces in entertainment platforms. As streaming services compete for audience attention, such innovations highlight AI’s role in transforming how consumers interact with digital content.
China’s Yizhuang Development Zone has unveiled subsidies worth tens of millions of U.S. dollars to build an end-to-end AI ecosystem running exclusively on domestic chips, operating systems and open-source frameworks. The goal: seed an ¥80 billion (US$11 billion) cluster by year-end, anchored on RISC-V processors and home-grown model libraries. Officials frame the plan as a self-reliance play, insulating critical industries from foreign export bans while cultivating local talent in high-performance GPU design. Observers note the timetable is aggressive: replacing Nvidia silicon and CUDA tools will require rapid breakthroughs in hardware and software compatibility. Still, the subsidies signal Beijing’s determination to blunt U.S. leverage over AI compute—and could spur a fresh wave of startup activity inside China’s borders.
🛠️ AI tools updates
Hugging Face has introduced a free agentic AI tool, designed to empower developers to create autonomous AI agents capable of performing complex tasks. This tool enables the chaining of AI models to tackle diverse applications, from scientific research to creative projects. By making it freely accessible, Hugging Face aims to democratize AI development, fostering innovation across industries. The tool’s versatility and ease of use position it as a game-changer for developers seeking to build scalable AI solutions without significant resource investment.
At its annual Sessions event, Stripe demoed an AI-powered “buyer agent” that can navigate checkout forms, authenticate payments and even batch-shop multiple blog recommendations—all through a single API call. Developers can spin up bespoke commerce agents in seconds, while Stripe’s new foundation model for transaction data aims to slash fraud and boost authorization rates. Future releases promise seller-side agents capable of negotiating prices in real time, hinting at web stores where bots do the haggling
💵 Venture Capital updates
Palo Alto-based Fastino landed a Khosla-led seed round to commercialize task-specific language models that run on low-end GPUs. The firm claims its architecture delivers single-token responses in milliseconds, undercutting big-model incumbents on both speed and cost.
Founded by Rubrik alumni, WisdomAI raised a Coatue-backed seed to build an analytics platform that writes micro-queries with generative models but leaves answer generation to deterministic data engines—sidestepping narrative fabrications. Early pilots include ConocoPhillips and Cisco.
🫡 Meme of the day

⭐️ Generative AI image of the day
Before you go, check out AI Unwraps a 2,000-Year-Old Vesuvius Scroll Without Touching It
