Technical.ly editor Katie Malone sends her Innovation Wire column as a newsletter every two weeks, covering local trends in AI, tech, startups, entrepreneurship and jobs of the future. Subscribe here.
Guardrails. Protections. Ethics. Governance.
Without proof behind it, “responsible AI” can sound like another vague and hard-to-define sales pitch. Vendors and public agencies often use the same language, many saying they will not buy or use AI unless it is safe, fair and accountable.
But who decides what that means and, even further, who checks the promise?
Across the country, frameworks for how governments use AI are popping up fast. States have task forces, acceptable-use policies, inventories, literacy plans and public principles. Cities and counties are writing rules, too.
AI literacy is useful, but it is not a substitute for enforceable standards governing how governments buy and deploy AI. I reviewed dozens of local policies to see where governments were getting more concrete. The strongest ones tended to ask for proof, not just promises.
Last week Mississippi Gov. Tate Reeves released the state’s first statewide AI education and workforce framework. The guidance says only that AI systems should be secure, monitored, auditable and subject to human oversight. A 2025 inventory by the state asked agencies whether an AI tool was reviewed for risk, security or privacy, but did not require that any oversight standards be made public, or even non-sensitive summaries of them.
Mississippi’s release instead largely talks about helping students, workers and the public understand the technology. That’s an area many states are concentrating on. North Dakota has an official K-12 AI Guidance Framework. There’s a Wisconsin action plan focused on helping workers gain AI skills, and a North Carolina executive order directing state officials to create public-facing AI literacy and fraud-prevention training.
Some states, including Pennsylvania, have created less formal public-facing AI resource hubs, aiming to give residents, families and educators one place to find guidance on using AI safely and responsibly.
A handful of jurisdictions are mandating specific oversight for government AI implementation.
In California, Culver City requires AI review, vendor documentation, contract addenda and named offices responsible for procurement, legal review and security. Seattle ties AI acquisition to review, possible impact assessment for high-risk systems and contract terms on bias, transparency, data use, auditing, human oversight and liability. Washington DC’s AI Procurement Handbook adds requirements around incident response, monitoring, bias, explainability, audit access and data handling.
Many, however, are doing a lot less. Some are still in the process of creating a task force to figure this out, while others set out their values, but no requirements on how to back that up.
None of these policies is perfect. Fact sheets can be seen as tiresome paperwork. Vendor answers can be thinly veiled marketing. Some security details should not be public.
It’s not that tough to write policies saying AI should be safe, fair, transparent and explainable. The test is whether those words turn into procurement standards: what vendors must submit, who owns the review, how often systems are reassessed and what happens when a tool fails.
Across the board, an unresolved issue is what happens after a first review. Without audits, enforcement and public accountability built in from the start, communities may still find out too late whether those guardrails actually worked.
Another angle:
“Compliance costs associated with regulation can further impede competition by disproportionately burdening smaller or newer firms that lack the resources of established competitors. High compliance expenses — including legal, administrative and technological requirements — may create significant barriers to entry, preventing potential competitors from entering the market or scaling effectively.”
Technical.ly headlines of the week
Stop re-prompting. Say it right the first time.
Voice-first prompts preserve the nuance you cut when typing. Speak once, paste into any AI tool, get results that don't need a follow-up. 89% of messages sent with zero edits.
3 things to know
💊 Rx-citing results: Five months into Utah’s AI prescription renewal pilot, early results are encouraging. Physicians agreed with the AI’s renewal recommendations 90% of the time. In about one-third of escalated cases, reviewers said the AI was being too cautious and could have handled the renewal itself. [TechTarget]
⏸️ Moratorium for thee, but not for me: Pulaski County, Arkansas, approved a one-year moratorium on new data center projects, but after a four-hour debate, officials allowed AVAIO Digital’s contentious project to keep moving — raising questions about what “paused” really means for projects already underway. [Arkansas Advocate]
💦 Smokey the Bear: Nothing beats fire prevention, but Jackson Hole, Wyoming-based Frontline Wildfire Defense says it can help reduce wildfire losses when flames get close. Its smart sprinkler system detects nearby fire threats and automatically soaks a home before windblown embers can ignite it, helping prevent catastrophe. [Cowboy State Daily]
What would you need to see before trusting your local government’s use of AI? Hit reply and let me know.