AI This Week: New Rules, New Models, and Cheaper APIs
Four stories define this week: Washington published its first serious attempt at federal AI law, Colorado rewrote its own AI statute with business-friendly changes, Google is days from launching its most powerful model yet, and the cost of running AI-powered tools has dropped to levels that make almost any automation project worth revisiting.
Congress Gets Serious About AI: The Great American AI Act Draft
On June 4, Representatives Jay Obernolte and Lori Trahan released a discussion draft of the Great American Artificial Intelligence Act of 2026 — the first serious attempt at a comprehensive federal AI framework in the U.S. The draft requires frontier AI developers to disclose technical details about their models, submit to third-party audits through designated Independent Verification Organizations (IVOs), and protect whistleblowers who report safety concerns.
This is a discussion draft, not a law — there’s no vote scheduled and the timeline to passage is unclear. But that’s exactly why this moment matters. Discussion drafts shape the final language, and the window for business voices to be heard is open right now. Industry groups, trade associations, and large enterprise users are already weighing in.
What this means for your business: If you build products on top of AI APIs, the audit and disclosure requirements being proposed here could eventually trickle down into contracts with AI vendors — or directly apply if you’re considered a “deployer” of consequential AI systems. It’s worth having a basic AI inventory now: what tools you use, what decisions they inform, and whether any touch hiring, lending, healthcare, or customer eligibility.
Colorado Rewrites Its AI Law — and Gives Businesses More Room
Just before the original June 30, 2026 deadline, Colorado Governor Polis signed SB 26-189 on May 14 — effectively replacing the state’s original AI Act with a substantially narrower law. The new compliance deadline is January 1, 2027, and the obligations are significantly lighter.
Gone are the sweeping risk management programs and impact assessments the original law required. In their place: targeted transparency and disclosure obligations focused on “automated decision-making technology.” The new law includes a 40-employee carve-out for small businesses, a 60-day right to cure for first violations, and a narrower focus on consumer disclosures. For any Houston business that has been quietly anxious about the Colorado law — especially companies using AI tools for hiring decisions, customer screening, or loan applications — this is meaningful relief.
What this means for your business: If you serve Colorado customers or employees and use AI tools in decisions that affect them, you now have until January 1, 2027 to comply — and the bar is lower than originally written. If you have fewer than 40 employees, you may be fully carved out. Either way, document what AI tools you use for consequential decisions. That step costs nothing and protects you regardless of where other states’ laws land.
Google Gemini 3.5 Pro Is About to Land — and It’s Built for Business
Google is days away from general availability of Gemini 3.5 Pro, with a launch window of June 23–30. The model ships with a 2-million-token context window — enough to ingest a year’s worth of business documents in a single request — plus a new “Deep Think” reasoning mode for complex problem-solving and strong multimodal capabilities across text, images, and data.
The competition this creates is good news for SMBs. With Claude Fable 5, Gemini 3.5 Pro, and GPT-5.6 all launching within a 10-day window, the pressure on pricing is real. Google is expected to price Pro at approximately $15 per million input tokens, competitive with Anthropic’s current rates. The 2-million-token context window is a practical tool for businesses dealing with long contracts, compliance filings, or large customer histories that previously couldn’t fit into a single AI session.
What this means for your business: If you’re evaluating AI for document-heavy workflows — reviewing vendor contracts, processing customer records, summarizing regulatory filings — Gemini 3.5 Pro’s context window and reasoning mode are worth testing when it launches this week. Google’s tight integration with Workspace (Docs, Gmail, Sheets) also makes it a natural fit for businesses already running on Google’s stack.
AI API Costs Have Fallen 40–70% This Year — the Math Has Changed
Buried under the model launch headlines is the story with the most immediate business impact: AI API costs dropped 40–70% across every major provider in 2026, with some models down more than 90% from their 2024 pricing. GPT-4o API access now runs $0.002 per 1,000 tokens — compared to $0.03 just two years ago.
The practical effect: multi-step AI agent workflows that cost $5–20 per run in 2024 can now run for under $1. SMB AI automation adoption has nearly doubled, from 22% to 38% between 2024 and 2026, and this pricing shift is a core reason why. Workflows that didn’t pencil out economically 18 months ago — automated customer intake, lead qualification, invoice processing, first-draft email responses — are now clearly cost-effective at even modest volume.
What this means for your business: If you looked at AI automation 12–18 months ago and the economics didn’t work, run the numbers again. A Houston service business handling 500 customer inquiries per month could run an AI triage layer for under $50/month at current API rates — a figure that makes almost any automation project worth a short pilot.
The takeaway
The AI story this week isn’t just new models — it’s the broader infrastructure starting to mature. Regulation is getting more specific and more workable. Costs are dropping. Capabilities are expanding. The businesses that act now, while that combination is in place, will have a measurable head start over those waiting for things to “settle.”
If you’re a Houston business trying to figure out which of this week’s changes affect you — and what’s actually worth doing before year-end — reach out to BlueHill. We work with SMBs to turn the week’s headlines into practical next steps.