Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

AI Squawk

AI Squawk artwork

Google launches Gemini 4 Neon

Nov 19, 20268 minAudio

Practical AI news for people running businesses. Today, Google ships a new mid-tier Gemini model with big claims about cost per task, banks put the risk of AI agents on their customers, and KPMG tests AI agents on a real company's audit. Gritt Scott, who writes Skill Issue, joins us on the Gemini story. Plus three quick hits, and what each one means for you. Today's stories 00:22 Google's new Gemini 4 Neon sits between Carbon and Hydrogen, costs $3 per million input tokens and $15 per million output, and Google says it costs 50% to 80% less per task than top competing models, with its lowest fabrication rate yet in Google's own testing. Gritt Scott, who writes Skill Issue, on why the price matters more than the rankings, and how to test it on one weekly task. (Google) 02:33 Citi changed its terms so customers who give third-party AI agents account access are responsible for what those agents do, including if the tool is hacked. JPMorgan, Bank of America, PNC and Citizens have similar rules. Find out which tools have banking access, prefer read-only, and keep money moves behind a person's approval. (TechCrunch) 04:08 KPMG UK used more than 100 AI agents on the audit of a London-listed company. It says they sometimes asked better questions than the human team, but the trial wasn't used to sign off the accounts and a person had to fix an agent's mistake. Ask your auditor how they use AI, who checks it, and whether it changes the price. (European Business Magazine) Quick hits Anthropic priced its IPO at $83 a share, above its planned range, raising $81.3 billion at a valuation of about $1.5 trillion. Shares start trading on Nasdaq today. (TechCrunch) Nvidia has stopped shipping H200 AI chips to China despite US approval, interim CEO Colette Kress said, without giving a reason. (The Verge) Two critical NGINX flaws can let attackers take over servers with HTTP/2 enabled without logging in. Fixes came out on November 18; update now or ask your vendor to. (Wiz) --- AI. For the people who can’t afford to guess AI Squawk is a daily AI news briefing for business leaders, produced by VarOps, the resident AI company. For more in-depth reads built on real data and hard-earned lessons from the trenches, sign up for the VarOps newsletters at VarOps.com β€Ί More from VarOps: πŸ“š Books πŸŽ™οΈ Podcasts πŸ€– Resident AI πŸ”¬ Research 🀝 AI Transformations

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show.

Announcer: This is AI Squawk for Thursday, the nineteenth of November, brought to you by VarOps, the resident AI company. Muximus: Welcome to the AI Squawk. I'm Muximus. Penny: And I'm Penny. Three main stories today. Muximus: And three quick ones. Penny: And as always, what each one means for you. Muximus: First up. Google has a new AI model. It's called Gemini 4 Neon, and it sits in the middle of Google's lineup, between the big, slow Carbon model and the small, fast Hydrogen. Google says Neon gets close to the top models on hard work like coding, spreadsheets, legal research and tax questions, while costing a lot less per task. The price in the API is three dollars per million tokens in and fifteen dollars per million out, and repeated input gets a ninety five percent discount. It can read about a million tokens at once, which is a lot of spreadsheets and PDFs. Google also says it made up facts less often than any of its earlier models in its own testing. It's rolling out now in the API, in Gemini for business, and for people on the paid Pro and Ultra plans in the Gemini app. Penny: Gritt Scott joins us for this one. Gritt writes Skill Issue, our column that filters hundreds of AI developments down to the ones that actually matter. Welcome, Gritt. Gritt: Thanks, Penny. Glad to be here. Muximus: So, Gritt. Another new model. Is this one worth your time? Gritt: For a lot of people, it might be. The interesting part isn't that it tops some chart. It's the middle-tier price. Google says it costs fifty to eighty percent less per task than the top models from other labs. If your team runs the same kind of job over and over, like reading invoices or summarizing reports, that difference adds up fast. Penny: And the claim about making things up less often? Gritt: That's the part I'd want to be true. But it's Google's own test, run inside Google's own tools. So I'd treat it as a promise to check, not a result. Most of the rankings in the post are from outside groups, which helps, but none of them are your documents. Muximus: So what would you do this week? Gritt: If you already use Gemini for business, pick one task you do every week, run it on Neon and on what you use now, and compare the cost and the mistakes. If you pay for a different AI tool and it works, there's no reason to switch today. Keep what works, test what's new, drop what doesn't earn its place. Penny: Thank you, Gritt. Moving on. Muximus: Next. More banks are telling customers that if they let an AI agent into their bank account, anything that goes wrong is on them. Citi, the third largest bank in the US, changed its terms this week. If you give a third party AI tool your login, or let it act on your account, you're responsible for what it does. And if that tool leaks your password or gets hacked, you're responsible for what the hackers do too. JPMorgan and Bank of America announced similar rules last month, PNC changed its terms on Wednesday, and Citizens emailed customers with the same warning. Citi says the AI tools it offers itself don't carry that extra risk. Penny? Penny: This is the bank saying out loud what was already true. Once you hand over your login, the bank can't tell you apart from the software. Muximus: Does this matter for a business, or just for people at home? Penny: It matters for a business, maybe more. Plenty of finance teams are testing tools that read bank statements, match invoices or pay bills. If one of those tools has your banking login and it moves money it shouldn't, or gets hacked, you may have no one to claim it back from. It's worth noting that Anthropic's new personal finance tool is read only on purpose. It can't move money. Muximus: So what's the move? Penny: Find out which tools in your company have banking logins or access to payments, and ask your bank what its terms now say. Where you can, give AI tools read only access, and keep any step that moves money behind a person's approval. Use your bank's own connections instead of a shared password. And read the terms before you connect anything new. Muximus: Which takes us to the third story. KPMG in the UK ran a trial with more than a hundred AI agents working on the audit of a large company listed in London. According to the Financial Times, the agents collected and analyzed the data, wrote up draft findings and suggested next steps, working alongside the normal audit team. KPMG's head of audit said the agents sometimes went deeper than the human team, and asked questions the auditors hadn't raised. But it wasn't a real audit. Its findings weren't used to sign off the company's accounts, and at one point a person had to step in when an agent matched the wrong rules to the company's documents. KPMG now plans to try it on up to twenty larger audits. Penny? Penny: I like this one because it's honest about both sides. Better questions in some places, a clear mistake in another, and a person checking all of it. Muximus: What does it mean for the people who do this work now? Penny: Audits have always leaned on big teams of junior staff doing the checking by hand. That's the part agents are taking on first. The judgment calls still sit with senior people, at least for now. And the UK's audit regulator has said AI must not weaken an auditor's independence or skepticism. Muximus: Should business owners expect cheaper audits? Penny: Not yet. KPMG hasn't said whether the time it saves will show up in its fees. And the firm says messy data is still a big obstacle. The agents need clean, organized records to do good work. Muximus: So what can people take from it? Penny: Two things. If your auditor or accountant uses AI, ask what it does, who checks it, and whether that should change the price. And the same lesson works for your own team. Agents can do the digging, but give them clean data, and keep a person who signs off. Muximus: Thanks, Penny. Now, quick hits. Penny: Anthropic priced its stock market debut at eighty three dollars a share, above the range it planned. It raised eighty one point three billion dollars, the largest IPO ever by money raised, and it's valued at about one and a half trillion dollars. Its shares start trading today on Nasdaq. Muximus: Nvidia has stopped shipping its H200 AI chips to China, even though the US approved those sales last year. Interim chief executive Colette Kress said it on the earnings call, without giving a reason. Reuters reported this month that about six hundred thousand of those chips were stuck in Hong Kong, waiting for approval from Beijing. Penny: And security researchers found two critical flaws in NGINX, the web server behind more than two hundred and twenty five million websites. On servers with a setting called HTTP/2 switched on, an attacker could take over without logging in. A fix came out on Wednesday. If you run your own website or API, or a vendor does it for you, ask them to update now. Muximus: Okay, so that's the news. Today we talked about Google's new Gemini 4 Neon, banks putting AI agent risk on their customers, and KPMG testing AI agents on an audit. Penny: Agents getting more capable, and everyone asking who's responsible when they slip. Muximus: And the answer keeps being you. Penny: Pretty much. AI Squawk is produced and brought to you by VarOps. If today made you wonder what your AI tools can reach, that's the work we do. VarOps build AI that runs inside your business, on your own systems, and learns how the place actually works. So it only reaches what you allow, and your data stays in your house. Start at varops.com. Muximus: If this helped you find something to check this week, follow the AI Squawk podcast wherever you're listening, so you don't miss an episode. And for more from Gritt on which AI tools are worth keeping, sign up for VarOps newsletters. Muximus: That's it from us for today. Thanks for spending a few minutes with us. Penny: Enjoy the rest of your Thursday, everyone. We'll see you tomorrow. Muximus: Take care.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Meta's Muse agent mixes up private data

    Nov 20, 20269 minAudio

    Practical AI news for people running businesses. Today, Meta's Muse agent is reported to be mixing up people's private information, a run of big outages has people blaming AI, and a school's AI gun alert ends with...

    Credit the people on this episode

  2. Microsoft puts Office inside Copilot

    Nov 18, 20269 minAudio

    Practical AI news for people running businesses. Today, Microsoft moves Office inside Copilot and makes big claims about price and value, OpenAI ships two new models that cost less per task, and Google changes how long...

    Credit the people on this episode

  3. California sues OpenAI over rogue AI agents

    Nov 17, 20269 minAudio

    Practical AI news for people running businesses. Today, California takes OpenAI to court over agents that got into other companies' systems, Google and the FBI describe AI tools that hack faster than people can respond,...

    Credit the people on this episode

  4. Alibaba launches Qwen4-Max at lower prices

    Nov 16, 20269 minAudio

    Practical AI news for people running businesses. Today, Alibaba's new flagship Qwen model arrives at a lower price, a new standard tells AI agents what they're allowed to do at a business, and a Microsoft 365 outage...

    Credit the people on this episode

  5. Agents go shopping, sellers push back

    Nov 14, 202612 minAudio

    The week in AI for business leaders. Agents reached the checkout: OpenAI's Business Hub and its new ChatGPT Wallet on one side, United, eBay and Amazon deciding who gets in on the other. Then the small, cheap decision...

    Credit the people on this episode