Google DeepMind Nears Gemini 4 Launch as Meta Bets Big on Muse at Connect 2025
Google DeepMind chief Koray Kavukcuoglu revealed that Gemini 4 is in its refinement stage and nearing launch, marking the company's first major flagship AI release in over 12 months, according to The Information. The announcement came during Kavukcuoglu's first media appearance since taking over the division, signaling Google's urgency to catch up with rivals like OpenAI and Anthropic who have shipped multiple frontier models in 2024 and 2025.
Meta CEO Mark Zuckerberg opened the company's annual Connect event in Menlo Park on Wednesday with a keynote focused heavily on Muse, Meta's AI agent platform, which the company says will reach over 500 million monthly active users across its family of apps, according to TechCrunch. Zuckerberg confirmed that Muse will also power features on Meta's AI glasses, deepening the integration between Meta's hardware and AI software ecosystems.
Meta also introduced camera-free AI glasses at the event, promising a lighter form factor with up to 12 hours of battery life, according to The Verge. The move signals Meta's strategy to expand wearable AI access while addressing privacy concerns that have shadowed camera-equipped smart glasses since their debut.
Experts warn that President Trump's framing of a US-China "AI race" could deter Beijing from sharing critical AI safety intelligence, potentially endangering US national security, according to The Washington Post. The analysis notes that over 60% of surveyed AI policy researchers believe cooperative frameworks are more effective than competitive posturing for managing catastrophic AI risks.
Tencent launched TenPayGo, a standalone payment app for foreign visitors to China that supports Visa, Mastercard, and Apple Pay across tens of millions of merchants, per Bloomberg. The product intensifies Tencent's rivalry with Ant Group and targets the roughly 30 million annual inbound travelers to China who have long struggled with digital payment friction.
XPRIZE announced winners of its $11 million Wildfire competition, where AI-powered detection systems spotted fires within 10 minutes but struggled to prevent their spread, according to Wired. The results highlight that early detection accuracy now exceeds suppression capability by a wide margin, exposing a gap between sensing technology and physical intervention infrastructure.
Sources: