Here is today's AI news. Story 1. OpenAI, Anthropic & Google DeepMind. The companies confirmed talks on an industry-led AI safety standards body, following a proposal by DeepMind CEO Demis Hassabis.. OpenAI confirmed talks with Anthropic and Google DeepMind on shared AI-safety standards, reportedly underway since July. The discussions followed Demis Hassabis’s proposal for a U.S.-led body modeled on FINRA. It could independently test frontier models before release, with industry funding and technical experts. OpenAI’s Chris Lehane said voluntary standards would complement—not replace—government safeguards. Story 2. OpenAI. OpenAI backed bipartisan U.S. congressional bills aimed at reducing the risk that AI models could help enable biological weapons and synthetic viruses.. OpenAI endorsed three bipartisan U.S. bills targeting AI-enabled biothreats: the Web of Biological Data Act, AI-Ready Bio-Data Standards Act and SCALE Biology Act. The measures would expand NIST’s role in biosecurity, standardize federally funded biological data for AI, and create an Energy Department data resource. OpenAI also supports a FRONTIER Act provision requiring independent evaluators to test leading models, potentially making biosecurity testing and external audits industry norms. Story 3. Google. Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, new models aimed at more natural real-time voice interactions and voice agents.. Google announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, new audio-to-audio models for real-time voice agents. Google says they enable “more natural, fluid, and intelligent” conversations, with background tool calls and visual context. The base model targets low-latency dialogue; Extended Thinking adds multi-step reasoning for complex tasks. Story 4. Meta. Meta launched Muse, a personal AI agent that runs on a secure cloud computer and can handle tasks such as email, travel booking, and form-filling with user approval for consequential actions.. Meta launched Muse, a personal AI agent initially available in the U.S. through a dedicated app, web, iOS, Android and WhatsApp. It can send emails, book travel, fill forms, make purchases and pursue longer-term goals, even in the background. Muse runs in an isolated “Muse Secure VM” with its own browser, while Sentinel controls network access and approvals. Meta says Muse cannot see passwords or payment details and pauses before sensitive actions. The launch intensifies competition in consumer AI while raising fresh concerns about privacy, prompt injection and autonomous access to personal accounts. Story 5. Anthropic. Anthropic launched Claude for Financial Advisors, connecting Claude to wealth-management and investment-analysis platforms including BlackRock, Addepar, and Schwab.. Anthropic launched **Claude for Financial Advisors** on Sept. 14, connecting Claude with platforms from BlackRock, Charles Schwab, Addepar and other wealth-technology providers. The tool supports meeting preparation, portfolio reviews, compliance checks and follow-up documentation. Its connectors can access Schwab balances, positions, transactions and alerts; BlackRock’s Advisor Center supplies model portfolios and analytics, while Addepar adds governed public- and private-market data. That's all for today's AI news briefing.