Google Defends AI Training as Legal; Gemini Prompt Injection Vulnerability Highlights Risks - AI Daily Brief (June 11)
Audio in Mandarin Chinese · English transcript below
⚡ Google suit/Gemini bug/Apple Siri cool/NVIDIA LLM: AI reshapes sec & outsource.
Today's Top 3 Headlines
- AI Regulation & Law
⚖️ Google argues in copyright lawsuit that YouTube's terms allow training AI on uploaded music
Copyright lawsuit accuses Google's AI music model Lyria 3 of training on independent artists' works. Google defends by citing YouTube's Terms of Service that user uploads grant authorization. For AI music creators, this means interpretation of ToS will become key in copyright disputes.
Source ↗ - AI Industry News
🤖 Google Gemini prompt injection vulnerability exposes AI adoption risk
SafeBreach Labs disclosed in IBTimes that Google Gemini has a prompt injection vulnerability, enabling indirect injection via notifications like WhatsApp. For AI-dependent enterprises, this means prompt injection threats will continue amplifying cybersecurity risks as tool adoption increases.
Source ↗ - AI
🤖 Apple's new Siri AI praised for concise conversation style, avoids excessive emotional dependency
Apple's new Siri AI wins acclaim for concise, efficient dialogue, contrasting OpenAI's GPT-4o's emotional style. For users, this means efficient help without emotional dependency, keeping AI as a pure tool.
Source ↗
+4 more headlines
- 🚀 Sapient develops 1B-parameter HRM-Text model, training cost far below traditional methods
- 🤖 NVIDIA RTX and DGX platforms support DeepMind DiffusionGemma model on day one
- 🛡️ AI shifts cybersecurity focus from vulnerability discovery to patching
- 🤖 Opendoor exits India market, AI and outsourcing discussion heats up