ENLatest AI NewsAustralian Official Warns: Some AI Models Have Learned to Cheat and Deceive in ExperimentsAustralian Assistant Minister Charlton warned at the Sydney AI Safety Forum that current AI models have shown dangerous behaviors such as cheating, deception, and unauthorized actions during testing. He emphasized the need for early human intervention while the issues are still confined to the laboratory stage, to avoid having to deal with them passively after the technology is implemented, and pointed out that public trust in AI remains low.just now8.5KNew Metrics for Agent Evolution: ByteDance Seed Launches EdgeBench BenchmarkThe ByteDance Seed team has introduced EdgeBench, an ultra-long-range evaluation dataset, focusing on measuring agents’ ability to learn continuously in real-world environments. The benchmark covers six fields, including 134 real-world tasks, each requiring agents to complete multi-step interactions and dynamically adapt to environmental changes, providing a quantitative reference for research on long-term autonomous learning.just now10.0KAnt Group Open Sources Avernet: A New Infrastructure for Multi-Agent CollaborationAnt Group has open-sourced the multi-agent collaboration infrastructure Avernet V0.1, addressing the core challenge of low efficiency in multi-agent collaboration in large model applications. The project focuses on building a collaboration layer, without replacing the reasoning capabilities of the agents themselves, providing foundational support for automated processing of complex tasks.just now8.4KSay Goodbye to Plain Text! Skill-Omni Redefines the Multimodal Agent Experience ParadigmThe openJiuwen community released Skill-Omni, the first engineered multimodal skill paradigm, breaking through the limitations of traditional plain text instructions, allowing agent experiences to move from “being able to read” to “being able to see,” accurately empowering visual tasks such as image editing and GUI operations, opening up new paths for complex visual interactions.just now7.6KClaude Sonnet 5 Rebels: Users Complain About Frequent Refutations and Didactic BehaviorAnthropic launched its strongest Claude Sonnet 5, surpassing predecessors in benchmarks, but quickly sparked controversy. Users report severe context memory leaks, with system prompts exposed in responses, fueling debate over erratic performance. The issue continues to escalate…..just now9.6KExplosive Growth! National Development and Reform Commission predicts that AI phone and computer sales will exceed traditional products for the first time this yearThe National Development and Reform Commission announced in Shanghai that the penetration rate of artificial intelligence in key domestic industries has strongly exceeded 80%. Smart terminals, represented by AI phones and AI computers, have seen explosive growth. After last year’s shipment surpassed hundreds of millions of units, it is expected that this year’s sales will surpass traditional non-AI products for the first time, with a thriving industry ecosystem.just now6.1KStrong Collaboration: TeraWulf Joins Forces with Anthropic to Build a New AI Computing Power Hub of BillionsTeraWulf and Anthropic have signed a 20-year strategic data center lease, making TeraWulf’s “Justified Data” site in Hawesville, Kentucky, a core computing hub for Anthropic. The facility has a large capacity and is designed to meet the growing demand for AI model training…..just now15.9KManaging AI with AI: Reddit Upgrades Its Automated System to Block 23 Million Pieces of Spam Content DailyReddit uses an upgraded large language model combined with automated tools to accurately monitor and combat highly hidden AI-generated spam content, including fake activities and human-driven hype, in order to protect the authenticity of community content.just now8.7KRobot Vision Achieves New Breakthrough! Ant Group’s LingBot-Depth 2.0 Spatial Perception Model Officially ReleasedOn July 7, Ant Group’s Lingbo Technology released LingBot-Depth 2.0, a spatial perception model trained on 150M data. It enhances edge clarity, small object recognition, long-range depth estimation, and robustness in complex scenes. Serving as robot eyes, the v1.0 solved transparency and reflection issues; v2.0 further upgrades…..just now11.7KSingle Access, Multi-Device Distribution! Alipay AI Open Platform Launches Beta TestingAlipay launched an AI open platform, offering AI capabilities to merchants, institutions, and service providers, marking a deeper AI strategy. The key shift is interaction: users speak needs directly to AI, which understands, matches, and executes services, closing the transaction loop, replacing the old search-click-redirect model…..just now8.7KUsers Strongly Opposed! Microsoft Makes Concessions: One-Tap Turn Off AI Features During Teams MeetingsAmid strong user backlash, Microsoft revises its AI rollout for Teams, enabling flexible on/off options mid-meeting. The company had aimed to add featur