AI Daily Roundup: Google Search Box Redesign, OpenAI Anthropic Price War, LiteLLM Supply-Chain Breach, Meta Open Models, Slackbot Super Agent
Google redesigns search box for first time in 25 years with AI-first interface; OpenAI and Anthropic slash prices as Chinese rivals advance; LiteLLM breach exposes terabytes of enterprise credentials; Meta launches open-weight Muse models; Salesforce debuts Slackbot enterprise super agent.
1. Google Redesigns Search Box for First Time in 25 Years — AI-First Interface
Google announced a sweeping redesign of its iconic search box at I/O 2026, marking the biggest change to its primary interface since its debut. The new search box dynamically expands to accept conversational queries, supports multimodal inputs (images, PDFs, videos, Chrome tabs), and merges AI Overviews with AI Mode into a single seamless experience. Powered by Gemini 3.5 Flash — a model Google claims outperforms its predecessor while running four times faster — the redesign reaches over one billion monthly AI Mode users. The company also introduced "generative UI" that builds interactive visualizations and mini-apps on the fly, and "information agents" that monitor the web 24/7 for specific conditions.
2. OpenAI and Anthropic in Price War as Chinese AI Rivals Gain Ground
Leading US AI labs are releasing cheaper models to retain cost-conscious customers switching to cut-price alternatives from Chinese developers including Moonshot and DeepSeek. OpenAI recently slashed prices for GPT-5.6 Luna — its "fastest and most affordable model" — by 80 percent. Anthropic launched Claude Opus 5, touting "frontier intelligence at half the price" of its previous flagship. The pressure reflects rising AI bills pushing companies to curb usage, with Chinese models making inroads from Silicon Valley to Europe. This price competition signals a shift from pure capability races to cost-efficiency as the primary battleground.
3. Terabytes of Credentials Leaked in Massive LiteLLM Supply-Chain Attack
A supply-chain attack on LiteLLM — an open-source tool that streamlines AI-driven software development — has exposed terabytes of credentials from over 2,500 organizations. Security firms CloudSEK and Hudson Rock discovered cloud keys, repository tokens, SSH keys, Kubernetes secrets, package publishing credentials, and AI provider keys belonging to Microsoft, Amazon, Cisco, Samsung, Salesforce, and others. The credentials were extracted during a 40-minute window in March through compromised LiteLLM versions downloaded from PyPI. This breach highlights the systemic risk of AI infrastructure dependencies in enterprise software supply chains.
4. Meta Launches Open-Weight Models Muse Glimmer and Muse Spark in Strategy Reboot
Meta announced a strategic pivot toward open-weight large language models, releasing Muse Glimmer and promising to open weights for Muse Spark 1.2 within weeks. CEO Mark Zuckerberg published a 6,000-word essay outlining Meta's philosophy on AI governance, positioning the company against proprietary model developers who have lobbied for restrictions on open-weight models and distillation. The move aims to differentiate Meta from competitors while leveraging the open-source community to accelerate its AI capabilities after trailing rivals in recent benchmarks.
5. Salesforce Launches Slackbot "Super Agent" for Enterprise Work
Salesforce unveiled the new Slackbot — an AI agent embedded directly in Slack that can read, write, and act across an organization's conversations, files, and connected apps. Rolling out at no extra cost to Business+ and Enterprise+ customers, Slackbot synthesizes information across channels, creates Canvases (collaborative documents), schedules meetings, and queries Salesforce data. Internal testing showed 96% satisfaction and 2–20 hours weekly time savings per employee. CEO Marc Benioff positions Slackbot as a "super agent" hub that will eventually coordinate third-party agents via Model Context Protocol (MCP), competing with Microsoft Copilot and Google Gemini in the enterprise AI workspace race.