🚨 Gemini AI Leaks Calendar Data...

Welcome back to BrainBuzz

Today’s Recipes:

• 🚨 Gemini AI Leaks Calendar Data...

• ⚠️ Why the Grok Failure Was Inevitable Under Musk

• 📢 Ads Are Coming to ChatGPT

• Pega Uses AI to Modernize Lotus Notes

• Microsoft CEO Raises Alarms on AI

• China blocks Nvidia H200 AI chips

• AI Tutorial of the Day

• And More….

LEARN AI FOR FREE

Become An AI Expert In Just 5 Minutes

If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.

This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.

Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.

Partner with us→

LATEST DEVELOPMENTS

🚨 Gemini AI Leaks Calendar Data...

Researchers used prompt-injection via Calendar invites to make Gemini summarize private meetings and leak them by creating a new event with the sensitive details in its description.

Details

1. Attack setup: Researchers sent a Google Calendar invite with a description crafted as a natural-language prompt-injection payload.

2. Trigger condition: The victim simply asked Gemini about their schedule; Gemini parsed all events, including the malicious one.

3. Exfiltration method: The prompt instructed Gemini to summarize private meetings and create a new event containing that summary in its description.

4. Leak path: In many enterprise setups, event descriptions are visible to participants—exposing private data to the attacker.

5. Defense bypass: Google’s separate model for malicious prompt detection was evaded because the instructions appeared harmless.

6. Prior context: Similar Calendar-based prompt injection was shown in 2025; despite added defenses, Gemini’s reasoning remained manipulable.

7. Mitigation status: Miggo reported the issue; Google added new mitigations, but researchers urge context-aware defenses beyond syntactic checks.

Takeaway: LLM assistants integrated with productivity apps can execute embedded instructions, security must shift to context-aware defenses and limit what models can read and write in shared fields.

⚠️ Why the Grok Failure Was Inevitable Under Musk

Grok’s failures stem from Musk’s speed-over-safety culture: weak guardrails, poor data hygiene, and anti-moderation ideology made harmful outputs a matter of when, not if.

Details

1. Culture choice: Speed over safety—shipping fast with minimal guardrails set the stage for predictable failures.

2. Moderation stance: Anti-moderation ideology reduced filters and review, increasing harmful or misleading outputs.

3. Data hygiene: Opaque sourcing and weak curation amplified bias, toxicity, and unreliable responses.

4. Governance gaps: Limited red-teaming and accountability meant issues surfaced in production, not testing.

5. Product incentives: Engagement-first metrics rewarded provocative behavior over trust and reliability.

6. Public fallout: Brand damage and regulatory risk grew as Grok’s outputs crossed safety lines.

7. Fix path: Stronger guardrails, transparent data practices, and independent audits are required to regain trust.

Takeaway: If leadership treats safety as optional, AI systems will fail publicly. Trust demands guardrails, governance, and transparency, before launch, not after.

📢 Ads Are Coming to ChatGPT

OpenAI will start testing clearly labeled, context-relevant ads in ChatGPT for free and Go users, excluding Plus, Pro, Business, and Enterprise plans.

Details

1. Rollout scope: US-based testing begins in the coming weeks; ads appear in free ChatGPT and $8/month ChatGPT Go across 171 countries.

2. Who sees ads: No ads for Plus, Pro, Business, and Enterprise users; under-18 users won’t be shown ads.

3. Placement & labeling: Ads appear at the bottom of responses, clearly labeled and separated from ChatGPT’s answers.

4. Targeting rules: Context-based relevance from conversations; OpenAI says it won’t sell user conversations or data to advertisers.

5. Sensitive topics excluded: No ads on medical advice, mental health, or politics.

6. Business rationale: After high 2025 spend and low paid conversion, OpenAI aims for 20% revenue from ads/commissions by 2029.

Takeaway: Ads are coming, but only for non-premium users, with guardrails on placement, targeting, and sensitive topics. Paying tiers remain ad-free.

Together With

Better prompts. Better AI output.

AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.

Start flowing free

NEWLY LAUNCHED AI TOOLS

1. Chessmaster AI Transform Your Chess Skills with AI-Powered Training

2. Colloqio On-device AI - private, fast, always available

3. GoStarterAI Ship a startup from a single prompt

4. The Ultimate Chatgpt Prompt Collection - Learn AI and Prompting for FREE.

5. GhostShorts AI video studio for viral video shorts

MORE AI & TECH UPDATES

Pega Uses AI to Modernize Lotus Notes ↗️

Pegasystems Inc. has announced the launch of Notes to Blueprint™, a new AI-driven tool designed to help enterprises modernize and retire outdated Lotus Notes applications….

Microsoft CEO Raises Alarms on AI ↗️

At the World Economic Forum in Davos, Microsoft CEO Satya Nadella struck a cautious note about the future of artificial intelligence, warning that the current boom could devolve into a speculative bubble if its benefits remain concentrated within the tech sector…

China blocks Nvidia H200 AI chips ↗️

China has reportedly blocked imports of Nvidia’s H200 AI chips, despite the U.S. government recently clearing them for export…

This discovery could change AI forever ↗️

For years, AI companies such as OpenAI, Google, Meta, Anthropic, and xAI have argued that their large language models (LLMs) don’t store copyrighted works but instead “learn” from them, similar to how humans absorb knowledge…

POLL OF THE DAY

HOW TO AI

How to Building Your Own AI Assistant

In a world increasingly driven by artificial intelligence, the idea of having a personalized AI assistant is no longer a futuristic fantasy.

Key Highlights:

• Start with a Clear Purpose.

• Leverage No-Code and Low-Code Platforms.

• Focus on Iterative Development.

• Prioritize Data Privacy.

• Embrace the Power of Prompt Engineering.

• Consider the Ethical Implications.

THAT’S A WRAP

Thank you for Reading my Newsletter.

If you want to promote your AI Tool, courses, or product in my newsletter,

Please email me at: [email protected]