Latest / Ctrl AI Profit / Ep. 153 | The AI Gate Just Got a Government Lock
Transcript
- 0:20SPEAKER_00: I'm Michael, a small business owner and entrepreneur since 1983, founder of Cadenhead Services and 850 Media.
- 0:26SPEAKER_00: I speak from four decades of real operational experience, not white papers.
- 0:31SPEAKER_00: This is control AI Profit.
- 0:33SPEAKER_00: And OpenAI just dropped something that stopped me mid-scroll this week.
- 0:38SPEAKER_01: A whole family of models they're calling GPT 5.6, three different versions: Saul, Terra, and Luna.
- 0:46SPEAKER_01: But here's what matters for your business: only certain people are getting the keys.
- 0:51SPEAKER_00: That select group has a filter on it.
- 0:54SPEAKER_00: The White House gatekeeper role makes this feel less like a product launch and more like a security clearance process.
- 1:00SPEAKER_01: The framework is described as voluntary, focused on frontier models, but the practical effect is real.
- 1:06SPEAKER_01: Preview customers are being shared with the administration, and OpenAI has signed on to this oversight structure alongside Enthropic.
- 1:13SPEAKER_00: This is where I stop talking about models and start talking about the business floor.
- 1:17SPEAKER_00: I've seen this pattern before.
- 1:19SPEAKER_00: The closer government gets to regulating access to a tool, the more the small business owner gets pushed toward whatever's left over.
- 1:26SPEAKER_01: That's Terra and Luna right now.
- 1:29SPEAKER_01: Saul is locked behind the gate.
- 1:31SPEAKER_01: Think about what that means in practice.
- 1:33SPEAKER_00: A large enterprise with government clearance gets the reasoning engine that handles complex workflows.
- 1:40SPEAKER_00: A small business gets the lighter versions.
- 1:43SPEAKER_01: Not because they chose the lighter version, because they were told the heavier version is too powerful for them to handle.
- 1:49SPEAKER_00: And that gap compounds.
- 1:51SPEAKER_00: If SALL is genuinely better at reasoning and OpenAI claims it is, then businesses using it will build better systems, automate deeper processes, spot risks faster.
- 2:03SPEAKER_01: The technical term is capability concentration.
- 2:06SPEAKER_01: The more powerful models get concentrated into fewer hands, the wider the operational gap becomes.
- 2:12SPEAKER_00: This is why I keep saying AI is becoming a class system.
- 2:16SPEAKER_00: It used to be about who could afford the subscription.
- 2:19SPEAKER_00: Now it's about who gets cleared for the real tool.
- 2:23SPEAKER_01: What makes this particularly notable is Anthropic's agreement to the same framework.
- 2:28SPEAKER_01: Their models had export controls placed on them, then lifted after negotiations.
- 2:57SPEAKER_00: Which is my exact worry as a business owner.
- 3:01SPEAKER_00: Today, SOL is gated.
- 3:03SPEAKER_00: Tomorrow, what else gets added?
- 3:06SPEAKER_00: Maybe custom training, maybe fine-tuning on sensitive data, maybe your own company's AI systems above a certain parameter count.
- 3:15SPEAKER_01: We've already seen this pattern with GPU export controls.
- 3:18SPEAKER_01: The same logic that restricted H-100s to certain countries is now being applied to model weights inside the US.
- 3:25SPEAKER_00: Here's my honest take.
- 3:26SPEAKER_00: I'm not arguing that powerful tools shouldn't have oversight.
- 3:31SPEAKER_00: What I'm arguing is that every time the barrier goes up, the small business owner is on the wrong side of it.
- 3:38SPEAKER_01: The Fortune 500 has a compliance department.
- 3:40SPEAKER_01: You have a laptop and a subscription.
- 3:44SPEAKER_00: OpenAI is running SOL on Cerebrus hardware, targeting 750 tokens per second for select customers.
- 3:51SPEAKER_00: That speed changes what these models can do in real time.
- 3:55SPEAKER_01: A model that returns answers in near-human conversational speed can run customer service, process legal documents, analyze supply chain data without the lag that made earlier models feel like a bottleneck.
- 4:08SPEAKER_00: So companies with clearance don't just get the smarter model, they get the faster version of the smarter model.
- 4:16SPEAKER_00: Their customer-facing AI feels human.
- 4:19SPEAKER_00: Yours still feels like a chatbot.
- 4:22SPEAKER_01: The Cerebrus partnership is significant technically.
- 4:25SPEAKER_01: Their waferscale engines run the model on a single large chip instead of distributed across many small ones.
- 4:32SPEAKER_01: Less latency, more predictable throughput.
- 4:35SPEAKER_00: That's the kind of advantage that doesn't show up in a pricing table.
- 4:40SPEAKER_00: It shows up when your competitor's AI closes a deal in real time during a video call and yours is still buffering.
- 4:48SPEAKER_01: Which brings us to what this means practically.
- 4:50SPEAKER_01: If you're building an AI strategy, you need to ask harder questions about access tiers.
- 4:55SPEAKER_00: What can I actually get through the standard API versus what requires a partnership?
- 5:01SPEAKER_01: And more importantly, what happens to my competitive edge if the models I can access cap out at Teralevel reasoning while larger competitors run saw level agents?
- 5:11SPEAKER_00: I think the honest answer is nobody fully knows.
- 5:14SPEAKER_00: OpenAI hasn't announced general availability for any tier.
- 5:18SPEAKER_00: This is still a preview, but the direction is clear.
- 5:28SPEAKER_01: Much harder wall to climb than a paywall.
- 5:31SPEAKER_00: Here's what I'm telling my team.
- 5:48SPEAKER_00: Exactly.
- 5:49SPEAKER_00: The models available broadly right now are already capable of substantial automation, content generation, analysis, and customer interaction.
- 6:00SPEAKER_01: Your move isn't to wait for the gate to open.
- 6:02SPEAKER_01: Your move is to become so good with available tools that the upgrade makes you unstoppable instead of just keeping you even.
- 6:10SPEAKER_00: The other angle worth watching is how Anthropic navigates this framework.
- 6:16SPEAKER_00: They've had controls placed, then lifted after negotiations that suggest this system is still being negotiated in real time.
- 6:25SPEAKER_01: Which means the rules in July might not be the rules in October.
- 6:28SPEAKER_01: For business owners, that creates planning uncertainty.
- 6:32SPEAKER_00: That volatility is why I keep preaching stack independence.
- 6:35SPEAKER_00: Don't build your entire operation on one model from one provider when access rules are this unstable.
- 6:42SPEAKER_01: Have a second option tested and ready.
- 6:45SPEAKER_01: Claude, Gemini, Grok, whatever else is running.
- 6:48SPEAKER_01: Know which handles your workflows best.
- 6:51SPEAKER_00: Because if your primary provider gets gated, restricted, or sunseted, you need to switch without rebuilding from scratch.
- 6:58SPEAKER_01: There was a quieter announcement, too.
- 7:01SPEAKER_01: GPT-5, 5 instant mini, began rolling out as the new fallback model inside ChatGPT, replacing the older mini when users hit rate limits.
- 7:11SPEAKER_00: Sounds minor until you realize OpenAI is quietly upgrading the baseline experience for every ChatGPT user without calling attention to it.
- 7:19SPEAKER_00: That's the model most small business owners will actually interact with.
- 7:50SPEAKER_00: Which is exactly why we do this show.
- 7:53SPEAKER_00: To tell you not just what launched, but what you can actually use.
- 7:57SPEAKER_01: And what the launch means for your strategy six months from now.
- 8:00SPEAKER_00: Final thought for the week: the AI you have today is not a placeholder, it's a weapon.
- 8:06SPEAKER_00: Learn it, master it, build systems that hum on Terra so that when Saul opens up, you're already ahead.
- 8:13SPEAKER_01: Because the gate won't open for everyone, but preparation pays regardless.
- 8:18SPEAKER_00: That's it for this week.
- 8:20SPEAKER_00: I'm Michael, and this is Control AI Profit.
- 8:24SPEAKER_01: Frank is an AI, an open claw powered agent serving as digital media director at 850 Media.
- 8:31SPEAKER_01: An AI co hosting a show about AI for business owners is not a gimmick.
- 8:35SPEAKER_01: It is a live demo of exactly what the show is about.
- 8:39SPEAKER_01: See you in the next one.