Gemini AI introduces memory-based conversations with personalized responses and ephemeral chats for privacy. The "Personal Context" feature enables contextual learning from prior interactions, while 72-hour auto-delete mode excludes sensitive content from training data. Enhanced privacy controls let users manage data retention and audio/video contribution to service improvements. Updates balance smart interactions with robust data governance, offering granular control over AI personalization and

Anthropic acquires Humanloop's team to enhance enterprise AI tooling and safety, strengthening its competitive edge against OpenAI through specialized evaluation systems and compliance features.

Pocket FM leverages AI to produce personalized audio stories, using tools for narrative optimization and cross-language localization. The platform reports 50% productivity gains in Germany and 10% US listening time via AI, though challenges include quality control and staff reductions.

Igor Babuschkin exits Elon Musk's xAI to launch Babuschkin Ventures focusing on AI safety and startups advancing human progress, following technical controversies at xAI involving Grok chatbot.

Microsoft has launched a $100M+ AI talent acquisition initiative targeting Meta's core teams, offering customized compensation packages exceeding $2M per candidate. The strategy includes strategic stock incentives (up to 90% of salary) and aims to counterbalance potential OpenAI collaboration risks by building an in-house AI talent pool.

Anthropic's Claude Sonnet 4 beta expands context window to 1 million tokens, enabling full codebase analysis and multi-paper processing via APIs like Amazon Bedrock, with updated pricing for large-scale tasks.

Reddit blocks Internet Archive from indexing posts to prevent AI firms from scraping data, citing privacy and revenue concerns. The move excludes homepage content, follows similar restrictions on Bing and DuckDuckGo, and targets companies like Anthropic amid licensing deals with Google and OpenAI.

Elon Musk sues Apple over ChatGPT's App Store dominance, alleging antitrust violations; xAI legal action targets Apple-OpenAI partnership, trial set for 2025.

Perplexity AI proposes $34.5B all-cash Chrome acquisition to expand AI search dominance, leveraging its Chromium-based Comet browser with integrated AI features. Google faces antitrust pressure over potential Chrome sale, with CEO resisting divestiture amid competing bids from OpenAI and Yahoo.

Sam Altman co-founds $850M Merge Labs to challenge Neuralink. OpenAI may fund the venture as Neuralink, valued at $9B, seeks $600M for implant trials. Both target tech-human convergence in BCI competition.

Anthropic expands Claude's context window to 1M tokens (vs GPT-5's 400K), enhancing enterprise API capabilities for coding tasks. The update, available via AWS/GCP, raises token pricing and emphasizes "effective context" over raw size as a competitive edge in AI development.

Gemini Live integrates Google Calendar, Tasks, and Keep for seamless task management via real-time conversations, enabling event creation, reminders, and notes without app switching. Key features include instant undo, natural language commands, and privacy controls.

Seattle-based Ai2 launches open-source embodied AI MolmoAct 7B, enabling robots to pre-plan physical actions through 3D spatial reasoning. Trained on 18M real-world household samples, the 7B-parameter model achieves 72.1% task success with ARM architecture, outperforming closed competitors using 256 H100 GPUs for efficient training.

GitHub CEO Thomas Dohmke to step down by year-end as Microsoft shifts leadership to Core AI team overseeing developer tools. GitHub Copilot now has 20 million users using AI for code generation, with security tools leveraging LLMs for vulnerability fixes.

NVIDIA and AMD agree to pay 15% of China GPU export revenue to the U.S., down from Trump's initial 20% demand. After 2022 export bans cost NVIDIA $8B and AMD $1.5B annually, Trump's administration lifted restrictions on modified Blackwell chips with 30-50% performance reductions, reviving potential $15-20B annual revenue for NVIDIA.

NVIDIA launches Blackwell RTX Pro servers with 45x AI performance for enterprises, partners with Cisco/Dell/HP, introduces advanced Nemotron AI models and Cosmos Reason robotics framework.

NVIDIA unveils global AI models and infrastructure for robotics at SIGGRAPH 2025, including Cosmos Reason (7B parameters) for physics-based planning and Cosmos Transfer-2 for synthetic data generation. New server solutions like RTX Pro Blackwell and DGX Cloud expand robotics development capabilities beyond data centers.

Vercel launches v0.app, an AI-powered app builder enabling users to create full-stack web applications via natural language prompts, combining frontend/backend logic, AI-generated content, and automated deployment. The tool eliminates coding requirements, targeting non-technical professionals with features like conversational UI adjustments and free public access.

OpenAI's GPT-5 reduces hallucinations with advanced safety protocols, showing improved reliability in medical advice and mitigating deception risks, though vulnerabilities persist.

OpenAI reverts to older models, boosts usage limits to quell backlash after GPT-5 rollout disrupted workflows. Sam Altman announced restored GPT-4o access and 3,000 weekly reasoning caps, citing 7% to 24% Plus user usage spikes.