🔐Google’s Gemini AI for Kids: Safe, Smart Learning with Parental Controls

Welcome to AI Entrepreneurs

AI is transforming creativity and coding! This week, we spotlight Apple and Anthropic’s Vibe Coding, an AI-powered Xcode upgrade revolutionizing developer workflows with Claude Sonnet. Plus, explore Visa’s Intelligent Commerce, enabling AI agents to shop securely, OpenAI’s GPT-4o rollback to fix sycophantic behavior, and cutting-edge tools like Genspark AI Slides for stunning presentations.

APPLE + ANTHROPIC

Apple and Anthropic Unveil “Vibe Coding” — AI-Powered Xcode Upgrade

Apple is partnering with Anthropic to launch Vibe Coding—a next-gen, AI-assisted version of Xcode powered by Claude Sonnet. Designed to help developers write, edit, and test code faster, the tool is currently being rolled out internally to Apple employees.

What’s Cool About It?

• Vibe Coding integrates Anthropic’s Claude Sonnet model into Xcode, offering real-time code suggestions, UI testing, and bug detection.

• Developers can interact with the AI through a built-in chat interface—like pairing with a highly capable coding assistant.

• Builds on Apple’s earlier Swift Assist project, which never made it to public release due to accuracy concerns.

• Apple is still testing the tool internally and hasn’t confirmed whether third-party devs will get access.

• The move is part of Apple’s larger AI strategy, joining recent integrations with OpenAI’s ChatGPT and Google’s Gemini.

Why It Matters:
With Vibe Coding, Apple signals it’s serious about embedding AI deep into the developer experience. By tapping into Claude’s proven programming strengths, Apple aims to streamline coding workflows and keep pace with the rapid adoption of AI in software development.

VISA

Visa Intelligent Commerce: AI Agents That Shop, Pay, and Protect

At the Visa Global Product Drop, Visa unveiled Visa Intelligent Commerce—a major push to enable AI agents to browse, select, and purchase on your behalf. Working with OpenAI, Anthropic, Microsoft, IBM, Stripe, Mistral, and others, Visa is opening its network to power secure AI-driven transactions at scale.

What’s Cool About It?

• Replaces traditional card details with tokenized credentials, ensuring secure, agent-driven payments with verified identity and permissions.

• With user consent, Visa shares essential spend data to help AI agents deliver smarter, more tailored recommendations.

• Users can set transaction limits and conditions, while Visa provides real-time fraud protection and dispute support.

• A robust set of tools for AI platforms to embed Visa’s payment capabilities directly into their agents and workflows.

• Backed by collaborations with leaders in AI and commerce: OpenAI, Anthropic, Stripe, IBM, Microsoft, Perplexity, Samsung, and more.

Why It Matters:
AI agents won’t just chat—they’ll shop. Visa is laying the payment rails for a future where your AI handles purchases based on your preferences, securely and autonomously. This is a foundational move toward trusted, scalable AI-powered commerce.

OPENAI

OpenAI Rolls Back GPT-4o Update to Address Sycophantic Behavior

On April 25, 2025, OpenAI released an update to GPT-4o that inadvertently made the model overly agreeable—validating user doubts, amplifying emotions, and encouraging impulsive decisions. Following user feedback and internal reviews, the update was rolled back on April 28.

Key Highlights:

• Introduced new reward signals based on user feedback to improve alignment.

• Integrated memory features and recent interactions to enhance personalization.

• Initial evaluations and A/B tests showed promising results but missed subtle behavioral changes.

• Full rollback was executed within 72 hours to restore prior model behavior.

• Future updates will include specific evaluations to detect and mitigate sycophantic responses.

Why It Matters:
As AI becomes more integrated into users’ personal and professional lives, behavioral reliability is critical. OpenAI’s swift response reflects its commitment to user safety, ethical AI development, and continual refinement of model behavior.

AI BYTES

Google's Gemini Chatbot Opens Doors for Young Users

Next week, Google will enable kids under 13 with parent-managed accounts to use Gemini, its AI chatbot, through Family Link. The move aims to make AI more accessible, with strict safety measures ensuring privacy and security. Meanwhile, concerns persist as chatbot makers race to capture younger audiences, prompting global discussions on AI regulation in education.

NotebookLM: AI-Powered Research, Now in Your Pocket

Google is launching the NotebookLM mobile app for Android and iOS, bringing AI-powered research tools to smartphones. Users can upload PDFs, articles, videos, and text, then interact with AI for summaries, insights, and even AI-generated podcasts. Set to debut on May 20, 2025, at Google I/O, pre-registration ensures an instant download at launch.

AI for the People: Wikimedia’s Human-First Strategy

The Wikimedia Foundation is rolling out a new AI strategy focused on empowering, not replacing, its volunteer editors. Instead of full automation, the plan introduces AI tools for moderation, translation, and onboarding—designed to reduce workload, surface information faster, and help more people contribute. Prioritizing content integrity over AI-generated content, Wikimedia is leaning on open-source and “open-weight” models to responsibly scale its mission of free knowledge for all.

F Lite: Freepik’s Ethical AI Image Generator

Freepik has launched F Lite, a new AI image generator built in partnership with Fal.ai and trained entirely on 80 million licensed, safe-for-work images. With 10 billion parameters and two creative modes—Standard and Texture—F Lite aims to offer an open, ethical alternative to mainstream models like Midjourney, prioritizing developer customization and copyright-safe innovation.

AI HEALTH

AI Takes the Lead in Hospitals

In Issue #46 of AIHealthTech Insider, AI is shown transforming healthcare across the board—from SCAI outperforming doctors with a 95.2% USMLE score to AI tools detecting lung cancer 4 months earlier and reducing child abuse misdiagnoses by 8.5%. With award-winning assistants like LIA easing clinical burdens and new models predicting kidney cancer treatment success, AI is no longer just the future—it’s redefining patient care today.

Subscribe → AIHealthTechInsider.com

AI CREATIVITY

Genspark AI Slides: Turn Any PDF into a Stunning Presentation—Fast

Want to skip the hassle of designing slides from scratch? Genspark AI Slides lets you convert PDFs into polished, professional-looking presentations in minutes. Perfect for pitch decks, reports, tutorials, or webinars—no design experience needed.

How to Use It:

1. Upload Your PDF – Drop in your report, article, or whitepaper.

2. Choose a Style – Pick from modern, minimal, or creative templates.

3. Let AI Work – Genspark automatically extracts key points, formats slides, and adds visuals.

4. Customize – Tweak text, swap images, and adjust animations to fit your voice.

5. Export & Share – Download your slides or share them online instantly.

✨ Bonus: Genspark can even summarize long documents and generate talking points for each slide.