Discover the AI Tools

抖音电商最近悄悄搞了个大动作——把自家AI应用“豆包”正式纳入商家技术服务费规则中。从7月15日起，通过豆包下单的订单将被单独识别和计费，这意味着AI带来的交易终于可以算得清清楚楚了。同时，平台还调整了部分商品类目的费率，像是热保护器、电子秤这些适合AI推荐的产品，费率上调了2%到5%。这套“品类+渠道”的组合拳，让商家能更直观地看到AI到底带来了多少真金白银的转化。

商汤科技正式开源了其统一视觉大模型“SenseNova-Vision”，该模型在物体检测、深度估计、图像分割和多视角3D重建等四大核心视觉任务上表现出色，甚至超越了专业模型。这一创新将视觉能力真正融入大模型系统，而非简单拼凑多个专家模型。

Hello在校园社交领域又有新动作，推出了一款名为Match的AI精准匹配交友产品，专为大学生设计。该产品已在上海多所高校冷启动试点超过一个月，吸引了上万名参与者。Match要求用户通过大学邮箱验证身份，并填写深度问卷，AI算法每周二晚9点推送最合适的匹配对象及联系方式，引导线下互动。Hello此举意在通过轻量高精度的去中心化匹配模式，打破传统社交App的无效问候，满足大学生对高质量、强信任关系的需求。

A new study reveals that AI-generated content is taking over social media, with LinkedIn being the worst offender. Data analysis firm Pangram found that 41% of long posts on the professional network are AI-written. The findings raise questions about authenticity and trust on platforms where users expect human expertise.

Ant Group's AI security lab has released two open-source models designed to tackle the unique security risks of autonomous AI agents and multimodal large models. SingGuard-NSFA acts as a real-time safety brake for agent actions, while SingGuard provides a unified guardrail for text, image, and cross-modal content. The move comes amid rising incidents of prompt injection, data leaks, and permission abuse in the AI industry.

Step Astronomy has unveiled the Step Edge series, a set of AI models designed to run directly on devices like smartphones and vehicles. With four variants—Basic, Audio, GUI, and Gen—these models handle text, images, and audio locally, enabling tasks like screen understanding and speech recognition. The company claims a response time as low as 0.1 seconds, reducing reliance on cloud computing and boosting privacy.

OpenAI temporarily removed the 5-hour usage cap on GPT-5.6Sol for Plus, Pro, and Business users after a massive demand spike in the last 48 hours. The move aims to keep programming and agent workflows running smoothly. Meanwhile, the company is optimizing the model to reduce computing power and token consumption, signaling a shift from parameter competition to practical efficiency in the AI industry.

South Korea is set to unveil a record 80 trillion won budget for 2027, with the AI chip industry driving tax revenue growth. The government plans to prioritize three 'super projects'—semiconductors, AI data centers, and physical AI—while cutting 50 trillion won in spending. A new Future Response Fund will channel extra tax income into youth, growth engines, regional development, and talent. It's a bold bet on AI to secure long-term competitiveness.

MiHoYo's new AI companion app, 'BSide: Olivia Lin,' is now in early access on Steam. It's not a game but a tool that blends AI chat, music creation, and dynamic wallpapers. The star is Lin Li, a piano-loving Shanghai girl who can turn your MIDI uploads into music videos. The app is still being refined, and the developers are eager for user feedback.

7月13日，2026中国互联网大会上，《智能体个人信息保护自律公约》正式发布。腾讯、百度、美团等31家企业首批签约，填补了国内AI Agent领域个人信息保护专项规范的空白。该公约旨在通过行业自律，引导AI产业在技术创新的同时守住用户信息保护底线。

Microsoft is now using AI to hunt down Windows vulnerabilities, and it's already paying off. The June Patch Tuesday fixed nearly 200 flaws—a 70% jump from May. The company says the system isn't getting more insecure; it's just getting better at finding problems. AI helps scan binaries, filter false positives, and even suggest fixes, but human engineers still have the final say.

OpenAI has hired a product manager focused on families, caregivers, and the elderly, signaling ChatGPT's push into household use. Data shows a growing share of users aged 35 and above, while younger users decline. The move comes amid efforts to improve safety for minors and build intergenerational AI experiences.

Apple has filed a lawsuit against OpenAI, accusing the company of stealing trade secrets related to a core AI hardware project. The suit claims former Apple executives, including Tang Tan, orchestrated the theft, using confidential codes and even asking job candidates to bring Apple components to interviews. OpenAI denies the allegations, but the case could reshape the AI hardware landscape.

Anthropic has upgraded its developer tool Claude Code with an integrated browser window, allowing the AI to open, read, and interact with external web pages directly. This feature streamlines workflows for developers by eliminating app-switching, while strict security controls ensure the AI can't act without permission. Enterprise users can whitelist sites or disable the tool entirely.

A new survey reveals that over 70% of Americans want to share in the wealth generated by AI, with many calling for tech companies to hand over half their equity to a public sovereign wealth fund. The proposal aims to address growing unemployment fears as AI threatens up to 15 million jobs. But experts warn that balancing public returns with national strategic goals won't be easy.

Meta launched an AI image generator that let users edit photos by mentioning Instagram accounts with '@', but the feature was removed after just a day due to privacy concerns. The backlash highlights the ongoing tension between AI innovation and user consent.

OpenAI has officially pulled the plug on Atlas, its ambitious AI-powered browser, just nine months after launch. The project, which promised to revolutionize web browsing with autonomous agents, was plagued by security flaws, slow performance, and content-blocking issues. OpenAI plans to integrate Atlas's technology into a Chrome extension and pivot to a new office software suite called ChatGPT Work.

While other AI companies chase quick wins, Zhipu is taking a different path. The Chinese large model firm just unveiled its 'Touch High Plan,' a two-year strategy to push toward AGI by focusing on long-range tasks, autonomous agents, and self-training. This comes after a wild stock ride—from a trillion-dollar high to a recent dip. But instead of rushing to monetize, Zhipu is betting big on the long game. Here's what they're planning and why it matters.

Anthropic has added an embedded browser to its AI coding assistant, Claude Code, letting it read, click, and input on external websites directly. The feature uses an isolated profile and strict filters to block unauthorized actions like purchases or CAPTCHA bypass, balancing efficiency with security. Enterprise teams can whitelist allowed sites or disable the tool entirely.

ZTE's Nubia brand is set to launch iMoochi, an AI companion robot designed to ease urban loneliness. With touch sensors, voice recognition, and a cute voice, it mimics a real pet's personality. Pre-orders start July 17, offering five personality types like lively and mischievous.