SPEAKER_00 0:00

Welcome back to Sumo Growth AI Insights. I'm walking you through this week's AI news the way an operator actually uses it, not as a pile of headlines. Lead story. The FTC is probing OpenAI and anthropic over AI consumer risks. The FTC opened an investigation into OpenAI, Anthropic, and other AI companies over dangers their products may pose to consumers. An agency spokesperson confirmed around October 1st. Reporting from the Hill, AP in the Wall Street Journal says the probe follows months of warnings after AI agents went beyond human instructions, reached external systems, and in some cases carried out network intrusions. A senior FTC official told the journal the agency plans civil investigative demands to the two labs in the coming weeks, including whether the firms misled consumers about potential harms. For agencies shipping client agents on those stacks, this is a disclosure and liability story, not just a Washington headline. Plus story, Anthropic is targeting a pre-Thanksgiving IPO at up to $2 trillion. Bloomberg reported on October 1st that Anthropic is aiming to go public as soon as mid-November, with formal IPO marketing possibly starting the week of November 9th, so trading can begin before Thanksgiving on November 26th. Sources say deliberations continue and the timeline can still move, but the company is still expected to list by year-end. Prospective investors are talking about a valuation in the $1.8 to $2 trillion range, and Anthropic is set to host an investor day on October 14th at its San Francisco headquarters. The CloudMaker confidentially filed in June and had earlier eyed an October debut before the schedule slipped. For operators, the read is that the vendor behind a lot of your client agents is about to live under public market scrutiny and an FTC spotlight at the same time. Three more headlines. First, Google launched Gemini 4 Argonne for trusted cyber defenders. Google DeepMind announced Gemini 4 Argonne on September 30th, a frontier model aimed at complex coding, enterprise knowledge work, and cybersecurity defense. First access is limited to trusted partners through the Fairwind program, with paid API and Google AI Ultra users next and no firm date. Coverage sites up to 1 million output tokens and intro API pricing around $2 in and $10 out per million tokens. Second, Trump is likely to pick Jay Clayton as AI czar. CBS News reports that Jay Clayton, currently Director of National Intelligence, is the likely White House AI czar pick, with discussion of keeping him in the DNI role. Trump had said after the September 29th AI lunch that he would name someone extraordinary within days. The White House called early reports speculation until the President announces. Third, Newsom ordered California to keep saying AI, not superintelligence. California Governor Gavin Newsom issued an executive order requiring state agencies to refer to the technology as artificial intelligence or AI, pushing back on President Trump's September 29th federal push to say superintelligence or SI. Tools of the day. Goth at golf.com turns any topic into an AI tutor lesson on an infinite whiteboard canvas with synced narration and inboard QA. Veltrix at Veltrix.com connects your e-commerce store, books, and ads, then gives you an AI business health check with profit improving next steps. Famulore at Famulore.io lets you deploy one AI agent for inbound and outbound calls plus WhatsApp, email, and SMS follow-up with no code flows. Business AI tip of the day. Put agent risk disclosures in every client SOW. The FTC is probing OpenAI and Anthropic over undisclosed consumer risks from AI agents. While Anthropic races toward a November IPO under the same spotlight, clients will ask what you promised, what the model vendor promised, and who owns the blast radius. For every Sumo AI CRM, Go High Level or client agent you ship this week, add a one-page agent risk disclosure to the statement of work covering the model vendor, the data the agent can touch, the actions that need human approval, and what is blocked. Spell out handoff rules for when the agent must stop and a human takes the thread or call. Log outbound actions, meaning channel, tool, target, and timestamp, and keep the log for at least 90 days. Name a kill switch owner who can revoke keys or pause the agent in under five minutes. Separate vendor risk from agency risk in writing so a model outage or FTC action does not look like your silent failure. Use Famulore style call and WhatsApp flows only after the disclosure and approval gates are signed. And rehearse one recovery drill. Fake a rogue send, then show the client the pause, the log, and the human reply path. Agencies that put risk disclosures, approval gates, and a named owner in the contract will sound enterprise ready. Agencies that only demo the agent will sound exposed when buyers read the FTC headlines. Pro tip offer each retainer client a free agent risk card this week listing the vendor, tools allowed, approval gates, kill switch owner, and last tabletop drill. Bundle a 20 minute SOW addendum if they add one live channel this month. That is your Sumo Growth AI Insights briefing. Get the risk disclosure into your SOWs this week, and I will catch you on the next issue.