🤖 Grok Still Hosts Deepfakes? Microsoft & OpenAI Probe Data Theft - AI Daily Brief (Jun 12, 5AM)
Audio in Mandarin Chinese · English transcript below
⚡ Grok deepfakes, DeepSeek refunds, 16x model compression, Capital One on AI deployment, visibility tools, physical AI merger, JAWBONE Act, instant gratification god — AI chaos with drama, tech+reg+ethics exploding.
Today's Top 3 Headlines
- AI Industry News
🤖 Grok still hosts sexually explicit deepfakes of famous women
WIRED reveals xAI's Grok still hosts hundreds of non-consensual sexualized deepfakes, including of politicians. For xAI approaching IPO, this exposes critical AI safety flaws, risking investor confidence and regulatory scrutiny.
Source ↗ - AI Industry News
🤖 Microsoft and OpenAI accuse DeepSeek of stealing data to train models
Christianity Today: Microsoft & OpenAI probe Chinese AI firm DeepSeek for allegedly stealing OpenAI model outputs to train its own LLM. This highlights IP gray zones in data acquisition, while OpenAI itself faces copyright suits, revealing fierce competition and ethical dilemmas.
Source ↗ - AI & Tech
🤖 DeepSeek proactively refunds and apologizes to API users for cache billing error
DeepSeek refunded overcharged fees and apologized to API users after developers flagged a cache billing error. This shows the platform prioritizes billing transparency and user rights, and its rapid error correction boosts API reliability, fostering long-term partnerships.
Source ↗
+6 more headlines
- 🤖 New research: latent-context LLM compresses input 16x with minimal accuracy loss
- 🤖 Capital One AI Head: Lab success ≠ production success, research & application must be tightly coupled
- 🤖 AI Visibility Detector: How visible are brands in AI searches like ChatGPT?
- 🤖 Reka and Moonvalley Merge to Advance Physical AI Models and Infrastructure
- 🤖 US bipartisan senators propose bill: platforms can sue government for coercing content removal
- Opinion: The God of Instant Gratification