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. Open AI, Anthropic, Google, and Meta testify under oath at a New York City AI hearing. Senior leaders from those four labs are set to testify under oath on October 5th before a rare New York City Council Committee of the whole, with all 51 members expected. Council Speaker Julie Menon said the labs must explain AI risks and weigh in on a package of about 10 proposed local AI bills after Anthropic, OpenAI, and Google agreed to appear only when the council threatened subpoenas. Named witnesses include Anthropic Frontier Red Team Head Logan Graham, OpenAI policy lead Morgan Dwyer, Google AI Policy Director Alice Friend, and Meta AI Legislation Director Shane Cahill, alongside whistleblowers Jacob Coxon, Alex Turner, and Daniel Kokotajlo. SpaceX AI was separately subpoenaed with possible court enforcement if it skips. For agencies shipping client agents on these stacks, this is a public record risk and disclosure story your buyers will quote back to you. Plus story, Sam Altman says the world should accept some AI harms. OpenAI CEO Sam Altman told Politico's decoded newsletter, reported October 4th by Reuters, that AI's benefits justify accepting some risks and that powerful models should stay broadly available. He said the world should accept some bad things happening for the benefits of this technology and people having the agency, and that there is still a lot of daylight between OpenAI and an anthropic on regulation. Altman called concentrating frontier AI in a single lab an unacceptable trade-off versus OpenAI's lighter touch stance and said he would not trade for zero hacks, misuse, or scams because he expects orders of magnitude more good than bad. Agencies selling OpenAI or Claude Stacks now have a clear quote to either own or distance from in every client risk conversation. Three more headlines. First, Softbank closed its about $3.1 billion digital bridge buy for AI infrastructure. SoftBank completed the acquisition of digital infrastructure investor Digital Bridge, announced in December 2025 and closed around September 30th. Digital Bridge stays a separately managed platform under CEO Mark Gonzi, with roughly $108 billion in data center, fiber, tower, and edge assets under management as Softbank scales AI compute. Second, Anthropic shipped Cloud India Inference on Amazon Bedrock. On October 5th, Anthropic said Cloud Opus 5, Sonnet 5, and Haiku 4.5 now run with in-country inference in India via Amazon Bedrock, so requests to the India Endpoint are processed inside India for regulated data residency. Third, OpenAI Safety Lead David Robinson quit over culture. Former OpenAI Safety Transparency Lead David Robinson resigned around October 3rd, writing that the company's culture is broken and the time for trial and error is over, and calling for nuclear and aviation style overlapping safeguards as models get more powerful. Tools of the day. FastRouter at fastrouter.ai routes each request to the right LLM across more than 200 models for cost, latency, quality, and failover. Spira Maxima at Spira.ai turns a script into a finished social video with presenter, B-roll, captions, and music in one pass. Invofox at Invofox.com extracts messy documents into SLA-backed structured JSON through one API. Business AI tip of the day publish your agency risk stance before clients ask. Frontier Labs are under oath in New York while Altman tells the world to accept some AI harms and an open AI safety lead walks out over trial and error culture. Your clients will not wait for a vendor FAQ. They will ask what you believe, what you disclose, and which stack you will still sell. For every sumo AI CRM, Go High Level, or client agent you ship this week, write a one-page agency risk stance covering which models you use, what harms you accept in writing, and what you refuse to automate without a human. Map each live agent to a named model vendor and a named approval owner. Add a public hearing or regulatory news clause to your statement of work so a New York City bill or FTC follow-up does not look like your silent failure. Keep a client-ready quote bank with Altman's risk acceptance, anthropic safety posture, and your agency's middle path in three bullets. Log every outbound agent action for 90 days so you can answer what it did the same day a buyer forwards a headline. Prefer routed stacks like FastRouter when you need failover and cost controls without hard coding a single lab and rehearse a five-minute client call script. What changed in the news, what is unchanged in your controls, and what you will update this week. Agencies that publish a clear risk stance sound like partners under scrutiny. Agencies that only demo the happy path sound like the vendors on the hot seat. Pro tip: send every retainer client a one-page risk stance card this week with vendor, approval gates, kill switch owner, and your written answer to Altman's Except Some Bad Things line. Bundle a 30 minute SOW refresh if they add one new agent channel this month. That is your Sumo Growth AI Insights briefing. Publish your risk stance this week, and I will catch you on the next issue.