Transcript Highlights:

• They're integrating weather – I mean, the radar models, many different sorts, in order to give predictions
• or any predictive modeling during a potential or an active emergency? No.
• What predictive modeling is out there?
• There is predictive modeling for some of the USGS gauges, NOAA, and the National Weather Service.
• They have gone in and modeled rainfall events to evaluate what the predictive elevation or flow will

Keywords: 997, house, all

Transcript Highlights:

• So we need a lot of data to make those predictions.
• So in all seriousness, what we do is we take those kinds of predictions and we take those kinds of predictions
• I don’t want my child’s data being used to train the AI models.
• All models are useful. All models are wrong. Some are more useful than others.
• is doing, and what the model is outputting.

Summary: The Committee on Education Pre-K through 12 held a panel discussion on artificial intelligence in K-12 education. Dr. Maya Israel of the University of Florida gave an overview of AI literacy, emphasizing that AI can support personalized learning, school operations, and teacher efficiency, but also raises concerns about data privacy, overreliance, mental health, and the need for human oversight. She described the Florida K-12 AI Task Force, which has produced guidance on policy, ethics, privacy, cybersecurity, classroom integration, and professional development, and is now working on district surveys, webinars, case studies, and teacher/family resources. Superintendents Van Ayers of Hillsborough County and Kevin Hendrick of Pinellas County described district-level implementation. Hillsborough outlined a governance council, a district AI policy, and an implementation guide; it prohibits generative AI for students below eighth grade, allows limited use for older students with teacher permission and district-vetted tools, and uses models for classroom AI expectations. The district also reported growing student enrollment in AI courses, teacher training through summer academies and quick learns, and pilots with tools such as Magic School and Microsoft Copilot. Pinellas emphasized academic integrity, student data privacy, digital responsibility, and the importance of human interaction in learning; it said state assessments already limit technology use and that districts need clear governance and parent transparency. Committee members asked about whether AI improves academic performance, the risk of student dependence, the possibility of returning to more paper-based testing, and who should set guardrails. Witnesses said reliable achievement data is still limited because the technology is new, though districts reported anecdotal gains in tutoring and early literacy. They also said there is no established statewide baseline for teacher AI training yet, so districts are using professional development, stipends, and partnerships with universities to build capacity. The committee also heard from Drew Andrew of FSU’s Inspire program, who argued that industry is moving faster than schools and that education should focus on teaching how AI works, building teacher confidence, and aligning training with workforce needs. No bills were considered and the meeting ended with a motion to adjourn, which was adopted.

Transcript Highlights:

• Population is still increasing, with a prediction topping out at 481 over the next three to five years
• They're just predicting how much population, and they're done at the top line.
• We could not predict. We knew we would have longer-sentence individuals.
• We have a decision package in for a staffing model.
• That was not a model that we had to do in JR.

Summary: The committee held a work session focused first on juvenile rehabilitation system capacity. DCYF officials said the juvenile rehabilitation population is older, includes more adult-sentenced youth, and has longer lengths of stay, especially for “post-25” youth who must remain in secure facilities and cannot go to community beds. They described overcrowding at Green Hill School, placement limits at Echo Glen and Harbor Heights, staffing turnover, mental health acuity, and the need for more medium-security and specialized mental health beds. DCYF said it is pursuing a Parkland facility proposal, a staffing model decision package, and a broader feasibility study and master plan update. No votes were taken; members were asked to follow up with questions later. The committee then heard on behavioral health system capacity from the Behavioral Health Administration and the Health Care Authority. DSHS described growth in forensic and civil bed need, expansion at Olympic Heritage, Maple Lane, and Brockman, and construction of a new 350-bed forensic hospital at Western State expected to open in 2028. HCA reported progress on long-term civil commitment beds, intensive behavioral health treatment facilities, PACT teams, and intensive residential treatment teams, saying the community-based system is being expanded to support step-down care and reduce hospital reliance. Members asked about whether capacity is right-sized, the difference between facility types, and federal match eligibility for services. A federal funding update followed, covering the effects of H.R. 1 and H.R. 5371 on SNAP, Medicaid, marketplace coverage, long-term services and supports, K-12, higher education, and hemp regulation. OFM and agency staff said H.R. 1 adds work requirements, changes non-citizen eligibility, increases state administrative and benefit costs, reduces Medicaid and marketplace subsidies for some groups, tightens redeterminations, and may significantly affect provider payments and state-directed payments. H.R. 5371 extended federal funding through January 30, 2026 and included some agency appropriations and other provisions, including changes affecting hemp producers. Members asked about SNAP error rates and special enrollment periods. Finally, budget coordinator Mary Monroe gave a 2026 supplemental budget preview. She reviewed the state’s near general fund outlook, noting revenue declines since the enacted budget, the effect of reversions, and a preliminary maintenance-level outlook showing a projected increase in NGFO spending over the four-year period. She said the supplemental will reflect updated caseload and cost forecasts and mandatory impacts from H.R. 1, but not policy proposals. No actions or votes were taken during the session.

Transcript Highlights:

• ,<00:29:20.399> we So creating the Minnesota model, we So creating the Minnesota model, we
• Interesting model and, uh, Mr.
• >> Yes, we are looking at this model as an opt-in model.
• The outdoor recreation endorsement model The outdoor recreation endorsement model creates<00:36:
• >> Yes, we are looking at this model um as an<00:37:02.160> optin<00:37:02.720> model

Bills: HF624

Keywords: HF624, Minnesota fishing law, game and fish, walleye stamp, walleye stamp validation, angling, angler, fishing lines, two lines, four lines, open water fishing, ice fishing, Department of Natural Resources, DNR, fishing regulations, fish harvest, conservation, Minnesota Statutes 97C.315, 1183, house

Transcript Highlights:

• Internationally, strong models like those in Vienna use public or nonprofit ownership, predictable funding
• Okay, some of the core assumptions of the models.
• No production model works at scale without the exemption.
• model.
• In British Columbia, the successful model we're operating, or excuse me, modeling under, has a program

Keywords: 988, house, all

Summary: The committee heard testimony on several housing-finance and permitting reforms aimed at making affordable and middle-income housing projects “pencil.” The first panel focused on the welfare property tax exemption, with witnesses arguing that annual recertifications are outdated, burdensome, and costly for both residents and operators. They urged streamlining by aligning eligibility rules with TCAC or HCD monitoring, allowing one-time qualification at occupancy, and preserving exemptions for projects that remain in compliance, especially as insurance costs and operating deficits are rising sharply. A major portion of the meeting centered on social housing and community land trusts under SB 555. HCD described the state’s ongoing study, due by December 31, 2026 and to be included in the 2027 annual report, and outlined public engagement already completed with residents and practitioners. Community land trust and policy witnesses argued that social housing will require legislative action beyond the study, including expanded tax abatements, public land use, soft loans, and simpler capital stacks. They emphasized that the model should include mixed-income and “missing middle” households, and several members discussed the stigma around the term “social housing,” suggesting a rebrand toward generational or multi-generational housing to broaden public support. The committee also discussed a proposed certified professional program modeled on Vancouver, Canada, to speed plan checks and inspections by allowing state-certified private professionals to perform certain code-compliance functions under local oversight. The witness said this would reduce delays, repeated reviews, and cost overruns while preserving local authority over zoning and enforcement. Members raised concerns about local control, infrastructure costs, and political resistance, but expressed interest in exploring a pilot and further recommendations. In the final panel, housing advocates supported allowing HCD loan funds to be disbursed during construction rather than only after completion. They said this would reduce interest costs, improve project feasibility, and could produce additional affordable homes without new appropriations. Members agreed the current system is fragmented and outdated, and several speakers and legislators repeatedly called for streamlined, more flexible financing and permitting tools to support housing production.

Transcript Highlights:

• You could, you know, better predict the price, but we predict that the price would be higher to achieve
• You could, you know, better predict the price, but we predict that the price would be higher to achieve
• You know, they have models, including the CATS models, and others, that are used to inform this.
• And that's really not true because every model is going to have assumptions required, and every model
• , and, you know, the models are uncertain and you have to account for that when you interpret modeling

Summary: The hearing was a select committee discussion on the transportation costs and impacts of California’s Low Carbon Fuel Standard (LCFS), with opening remarks from the co-chairs and members emphasizing affordability, climate goals, and the need to explain the program’s benefits to the public. The first panel from CARB and the California Energy Commission described how LCFS works as a market-based, declining carbon-intensity program that rewards lower-carbon fuels, supports zero-emission vehicle infrastructure, and is intended to reduce greenhouse gases and local air pollution. They argued the program has driven billions in private investment, increased alternative fuel use, and that LCFS credit prices are not the main driver of retail gasoline prices, which they said are dominated by crude oil, refining, and distribution costs. Members questioned the panel on the gap between the regulatory target and actual carbon-intensity performance, the role of credit banking, which fuels are generating the most credits, how the 2025 amendments affected the program, and whether LCFS credits are truly additional. CARB explained that banking helps cost-effectiveness and investment certainty, that ethanol, renewable diesel, and biodiesel currently provide the largest volumes while electricity is expected to grow, and that the updated targets were informed by the state’s 2045 carbon-neutrality goals and the 2030 scoping plan. The Energy Commission said its data show environmental programs add some cost to gasoline but do not drive price volatility, which is mainly tied to crude oil and refinery margins. The second panel, featuring academic and research experts, focused on program design, out-of-state credit generation, and broader economic effects. Speakers said LCFS is successful because it ties incentives to emissions benefit, uses life-cycle analysis, and allows flexible compliance that lowers costs compared with more direct regulation. They also said the program’s benefits generally outweigh costs, that it can reduce air pollution disparities and support equity, but that some issues—especially indirect land use change, additionality, and older program assumptions—need more research and may warrant future rulemaking. One researcher noted that while LCFS likely raises gasoline prices somewhat, the effect is uncertain by design and usually smaller than normal market fluctuations, and another warned that limiting credit generation too narrowly could create legal and efficiency problems.

Transcript Highlights:

• They create probability models and they start to make predictions.
• And over time, those machines get smarter at making predictions.
• So open models and other models out there often times people are wondering around accuracy. Right?
• tracks evaluations across these model.
• We've launched I age prediction that goes beyond verification of predicts the age of the user.

Transcript Highlights:

• We know the data that we used, the methods that we use, the model that we use.
• And modeling those COLAs is one of our key purposes of this assumption.
• And modeling those coales is one of our key purposes of this assumption.
• All economists I see are predicting inflation to go up because of tariffs.
• We put those together to model total salary growth for each member in the system.

Summary: The committee approved the July minutes and then received an informational presentation from the Office of the State Actuary on the financial condition of the state retirement systems. The actuary reported that employer contribution rates are generally declining, helped by strong investment returns and reduced funding for PERS 1 and TERS 1, while funded ratios have continued to improve; on a combined basis the plans were reported at 100% funded in 2024, with open plans above 95% and legacy plans varying by system. The presentation also reviewed projected rates and funded ratios under current assumptions, noted that pension costs are taking a smaller share of the state general fund, and discussed risks from investment volatility, policy changes, and demographic experience. Committee members asked about savings from lower rates, deferred asset smoothing, and how Washington compares with other states. The committee then considered the state actuary’s recommendation on long-term economic assumptions and adopted all four recommendations by roll call votes: inflation at 3.0%, general salary growth at 3.5%, membership growth for Plan 1 funding at 1.0%, and investment rate of return at 7.25%. The actuaries explained that the inflation and salary growth increases were driven largely by higher long-term inflation expectations, while the investment return recommendation matched the current statutory assumption. Members discussed the timing of the Pension Funding Council’s decision, the effect of tariffs and inflation uncertainty, and how assumption changes would affect future contribution rates and budgets, particularly for open plans. Staff then gave an update on the LEOFF 1 study, explaining the difference between being “ahead of schedule” and truly overfunded, and summarizing responses received from DRS, the State Treasurer, and the State Investment Board on the merger and restatement proposals. DRS said both bills could be administered, though the merger bill’s COLA banking provision would be challenging until its new system is ready; the Treasurer urged caution, especially about the restatement bill and the use of one-time funds; and the Investment Board said removing assets from the trust would have some transaction costs but likely small impacts. The committee discussed whether to invite additional agencies and local government groups to testify, and staff said more responses, including from Ice Miller and the State Actuary, were expected for the October meeting. Finally, the committee heard a briefing on PERS 1/TERS 1 COLA policy and related bills from the last session. Staff reviewed the committee’s prior ongoing COLA recommendation, the SCPP-endorsed bills that would have created a one-time 3% COLA followed by an ongoing COLA, the Senate merger bill, and a separate ad hoc COLA bill. Public testimony largely supported Plan 1 COLAs and stable contribution rates, while several speakers urged caution about transferring LEOFF 1 surplus assets or merging legacy plans, and others raised concerns about climate risk and the pension fund’s investments. No further committee action was taken on the COLA item during this portion of the meeting.

Transcript Highlights:

• EARLY 2020S WE REALLY TRANSITIONED SO THINK CHAT GPT, THAT'S A LINE EXAMPLE OF THE LARGE LANGUAGE MODEL
• ON THE CLINICAL SIDE WE ARE USING A PREDICTIVE TOOL THAT HELPS US WITH EARLY DETECTION OF SEPSIS.
• THESE ARE NOT PUBLIC AI MODELS, THEY LEVERAGE INSIDE DCF CLOUD INFRASTRUCTURE.
• WE NEED TO KNOW ONE OF THE DATA THAT ARE USED TO GENERATE THE PREDICTED AI MODELS.
• MODEL OF THE CLINICAL SETTING IS DIFFERENT THAN ME AS A HUMAN MAKING CLINICAL JUDGMENT.

Transcript Highlights:

• We utilize a combination of time series forecasting and discrete simulation modeling to project Predict
• Like, what is the predictive form? What is...
• modeling.
• So these models don't assume any... changes in those numbers.
• We have a unique health care model, I think a very good model, with the universities.

Keywords: 1184, house, all

Transcript Highlights:

• Again, these are models; these are not actuals. It is really hard to get actuals here.
• It's a model actually across the country as well.
• And unfortunately, housing, it's an outdated model as someone who's been in that space.
• Predictability and clarity, not just new programs, unlock financing.
• But how can we legislate to create a more predictable entitlement process?

Summary: The Assembly Select Committee on Housing Finance and Affordability held its first hearing of 2025 to examine California’s housing finance system, with opening remarks emphasizing the state’s severe housing shortage, high costs, and the need for practical recommendations to the Legislature and Governor. Co-chairs described the committee as an educational and problem-solving forum focused on financing housing production, first-time homeownership, mixed-income developments, and affordability across the income spectrum. Witnesses from state agencies and the development sector were invited to explain how housing is financed and where the system is breaking down. Panelists from the California Housing Partnership, the Business, Consumer Services and Housing Agency, the Tax Credit Allocation Committee/State Treasurer’s Office, CalHFA, and Related outlined the “capital stack” used to finance affordable housing, stressing that projects typically rely on multiple public and private sources, including federal and state low-income housing tax credits, tax-exempt bonds, state subsidies, local funds, and rental income. Speakers noted that affordable housing rents generally cannot support full project costs without public subsidy, and that recent federal changes—especially the expansion of the 4% and 9% tax credit programs and the reduction of the bond financing threshold for 4% credits—should allow California to finance substantially more units. CalHFA also described its homeownership programs, including My Home, Dream For All, and disaster-related mortgage assistance, as well as its multifamily lending and bond issuance programs. Several witnesses and committee members emphasized that the system remains too complex, too slow, and underfunded. They pointed to the need for more state funding, a housing bond, a permanent funding source, and better coordination among agencies, while also citing recent streamlining efforts such as AB 434’s SuperNOFA, AB 519’s one-stop-shop working group, and the planned California Housing and Homeless Agency reorganization. Members raised concerns about equity, access, missing-middle housing, gender and racial disparities, and whether current programs adequately serve extremely low-income households and those at risk of homelessness. No formal votes or actions were taken during the hearing; the discussion ended with committee members and witnesses agreeing that both funding and administrative reform are needed to increase production and improve affordability.

Transcript Highlights:

• It was bold, it was historic, it was a national model.
• It was bold, it was historic, it was a national model.
• The bill makes reimbursement for homeschool families more predictable and parent-friendly.
• We've been the model for everything that we've done to be able to provide choice for parents.
• So I predict to you, Mr. Chairman, that this will be not the last time. So I predict to you, Mr.

Summary: The Senate Appropriations Committee met with a quorum present and took up two bills. The first, SB 250 on rural communities by Senator Simon, was described as a broad rural development package creating an Office of Rural Prosperity, a Renaissance grant program, housing and transportation investments, added funding for rural education consortiums, and health care initiatives for rural areas. Senator Harrell asked about overlap between road funding programs, and the sponsor explained that eligible counties could receive both SCRAP and FARM funding. Several organizations waived in support or spoke in support, and the bill was reported favorably by unanimous vote. The committee then heard SB 318, the committee substitute for educational scholarship programs by President Gates. The bill was presented as a response to Auditor General findings about the rapid growth and administration of Florida’s school choice and scholarship programs. It would separate Family Empowerment Scholarship funding from the FEFP, require more frequent student enrollment verification, lower scholarship funding organization administrative fees, require return of overpayments, create a student ID system, establish a $250 million stabilization fund, require annual audits, and direct DOE to recommend future program administration through competitive procurement. Gates also offered five amendments, including technical changes to eligibility documentation and a substantive amendment requiring a DOE report on future administration and competitive selection; all five amendments were adopted. During debate and public testimony, senators and witnesses discussed accountability, software solutions, reimbursement delays, monthly attestations, and impacts on public schools and families. Supporters and opponents alike raised concerns about bureaucracy, fraud prevention, special education services, and whether the bill would help or burden parents. Gates said the bill aimed to fix tracking and payment problems without capping the program, and he noted the IEP timeline would be aligned with public school timelines. After debate, the committee reported CS for SB 318 favorably by unanimous vote, and then adjourned.

Transcript Highlights:

• Invest in localized predictive weather modeling to better serve rural communities.
• Okay, and in so doing with that, to project to predict a flood would require a model, which is being
• Regarding better data and models, I'll focus on models of the atmosphere that predict precipitation.
• , I'll focus on models of the atmosphere that predict precipitation.
• The other is using the numerical models that will assist you in making future predictions of what maybe

Summary: The joint Senate and House disaster preparedness and flooding hearing opened with quorum calls, prayer, testimony time limits, and remarks from legislative leaders emphasizing decorum, respect, and the goal of learning from the July 4 flood to prevent future loss of life. Lieutenant Governor Dan Patrick and Speaker Dustin Burrows said the special session and hearings were intended to hear directly from local officials and residents, and that the state would pursue immediate and longer-term legislation and continue working beyond the current session. The committee then heard invited testimony from Kerr County, Kerrville, Kendall County, Real County, and the Upper Guadalupe River Authority, followed by member questions. Kerr County Judge Rob Kelly, Sheriff Larry Leitha, and emergency management coordinator William B. Thomas described the flood as a sudden, unprecedented event that overwhelmed local systems before many officials or residents understood the danger. They said the county received multiple National Weather Service and IPAWS alerts, but the rain fell in remote upstream areas with limited gauges, and the flood rose too quickly for normal response. They detailed the emergency response, including dispatchers handling heavy call volumes, deputies and volunteer firefighters conducting rescues, and state and federal partners assisting. County officials said 108 people died, including 37 children, and two remained missing. They called for better real-time flood detection, stronger rural emergency management resources, improved broadband and cell coverage, and more funding for swift-water rescue and warning systems. Kerrville Mayor Joe Herring and City Manager Dalton Rice said the city had floodplain rules, emergency plans, and prior preparedness efforts, but the event exposed gaps in communications and regional warning capacity. They urged the Legislature to fund a flood warning system, whether sirens, automated alerts, or a hybrid, and asked for support for predictive weather modeling, broadband expansion, a hazard mitigation fund, and disaster recovery resources. The Upper Guadalupe River Authority said it had long funded gauges and pursued flood-warning and mitigation projects, including a new software-based flood prediction system and possible retention dams, and planned to dedicate reserve funds to feasibility studies and infrastructure improvements. Kendall County Judge Shane Stolarczyk said his county’s emergency action plan and early alerts helped prevent deaths there, and he recommended easier access to alerting tools, a river-wide warning system, simpler grant applications, and regional coordination. Real County Judge Bella Rubio said rural counties face major challenges because of limited cell service, broadband, staffing, and funding, especially for camps and visitors along the Frio and Nueces rivers, and argued that reliable alert systems should be treated as a necessity, not a luxury. Members asked about the timeline of the flood, low-water crossings, communications failures, sirens, bridges, and funding, and several witnesses said the key lessons were better communications, better warning systems, and more investment in rural flood mitigation and emergency response.

Transcript Highlights:

• Invest in localized predictive weather modeling. to better serve rural communities.
• Predict it with 75% confidence that the flood is going to happen based on early predictions. then upgrade
• Regarding better data and models, I'll focus on models of the atmosphere that predict precipitation.
• They take the NWS model models, and forecasts, and apply them specifically to customer needs.
• The other is using the numerical models that will assist you in making future predictions. of what maybe

Keywords: 1185, senate, all

Transcript Highlights:

• legislation on prediction markets and Legislation on prediction markets, and you guys really came up
• So Hawaii is a model for the country.
• So Hawaii is a model for institution. So Hawaii is a model for the<01:08:47.199> country.
• Keep in mind that<01:10:07.199> FanDuel that FanDuel that FanDuel predicts, predicts, predicts
• , DraftKings DraftKings DraftKings predictions, predictions, predictions, and<01:10:15.040> Fanatics

Bills: HB2227, HB1775, HB1560, HB1642, HB1647, HB2003, HB1640, HB2198

Keywords: eviction, housing stability, tenant screening, court records, writ of possession, foreclosures, public sale, eligible bidders, housing, downpayment, judicial foreclosure, digital financial assets, consumer protection, financial kiosks, refunds, transaction security, HB1642, Hawaii, cryptocurrency ATM, crypto ATM

Summary: The committee on Consumer Protection and Commerce met on February 5, 2026, and heard testimony on several bills, beginning with HB 227 relating to eviction records. Supporters, including the Public First Law Center and the Office of Hawaiian Affairs, argued the bill would help people who prevail in eviction cases avoid long-term housing harm from online court records, while the Public First Law Center said keeping records off eCourt Kokua would not violate the First Amendment because the records would still be available in person. Members discussed access-to-justice concerns, and a witness said legal aid attorneys could still access the records through the attorney-only Jeff’s system and the court’s access-to-justice room. The chair also asked about precedent, and a witness cited a Hawaii Supreme Court case as supporting removal from the online database rather than sealing records entirely. The committee then took up HP 1775 relating to foreclosures, but the transcript only shows in-person opposition comments from the Hawaii State Bar Association Collection Law Section, the Hawaii Credit Union League, and the Hawaii Bankers Association. The credit union and banking groups said they had concerns about broader negative impacts on mortgage lending and other requirements, but no detailed discussion or action was captured before the committee moved on. The next measure, HB 1560 relating to consumer protection, drew support from the Office of Consumer Protection and cryptocurrency companies including Coinflip and America Digital, which said they already use wallet-pinning and other safeguards to prevent fraud. AARP Hawaii did not take a formal position but said the bill addressed a real problem, noting that Hawaii residents, especially in Kona, had lost more than $920,000 in 2024 to cryptocurrency ATM scams and arguing that stronger oversight was needed. The committee also heard HB 1642, which would ban cryptocurrency kiosks. The Office of Consumer Protection supported the ban as the best way to protect consumers from fraud, while Coinflip, Bitcoin Depot, and America Digital opposed it, arguing kiosks provide cash-based access to crypto, especially for unbanked or underbanked consumers, and that targeted regulation would be better than an outright ban. AARP Hawaii took no formal position but strongly emphasized the harm caused by scams, saying victims are often frightened into acting quickly and that kiosk transactions currently lack enough friction or intervention. Finally, HB 1647, also on consumer protection, would impose liability on host businesses that provide space for crypto kiosks. The Office of Consumer Protection warned small businesses might not understand the liability, while Coinflip, Bitcoin Depot, and America Digital opposed the bill, saying it would unfairly shift enforcement duties to host stores and could discourage businesses from hosting kiosks, effectively creating a de facto ban. No votes or final committee actions were taken in the portion of the meeting provided.

Transcript Highlights:

• We need stable, predictable funding to meet our needs.
• <00:56:11.400> funding raise critical and predictable funding raise critical and predictable
• Real results will not be seen without real, stable, and predictable money.
• represented finally predictable represented finally predictable dedicated<01:02:38.480> Revenue
• without real stable and predictable without real stable and predictable money<01:02:51.640> it

Keywords: 1187, senate, all

Transcript Highlights:

• Invest in localized predictive weather modeling to better serve rural communities.
• To predict a flood would require a model, which is being developed, I understand.
• Regarding better data and models, I'll focus on models of the atmosphere that predict precipitation.
• The other is using the numerical models that will assist you in making future predictions of what the
• A prototype of this model was developed in Austin at the Texas Advanced A prototype of this model was

Summary: The joint Senate and House disaster preparedness hearing convened in Kerrville with quorum, public testimony limited to three minutes and invited testimony to 10 minutes. Leaders from both chambers, along with the lieutenant governor and speaker, framed the hearing as an unprecedented joint effort focused on learning from the July 4 flood, honoring victims, and identifying actions to reduce future loss of life. The committee also heard opening remarks about decorum, logistics, and the intent to continue work in future sessions. The first panel included Kerr County Judge Rob Kelly, Sheriff Larry Leitha, Emergency Management Coordinator William B. Thomas IV, Kerrville Mayor Joe Herring Jr., Kerrville City Manager Dalton Rice, Upper Guadalupe River Authority representative William Rector, Kendall County Judge Shane Stolarczyk, and Real County Judge Bella Rubio. They described the flood as sudden and catastrophic, with Kerr County reporting 108 deaths and two missing. Local officials emphasized that they received no timely warning of the scale of the event, that responders and volunteers acted heroically under extreme conditions, and that communications, cell coverage, and rural emergency resources were strained. Several witnesses said the county’s existing alert systems were limited by geography, sparse broadband, and the speed of the flood. Testimony focused on possible improvements, including real-time flood gauges and predictive monitoring, stronger rural emergency management staffing and training, better interoperability and alerting tools such as IPAWS, CodeRED, WENS, and sirens, and expanded broadband and radio coverage. Kerrville asked for a flood warning system before next summer and state help for stormwater, floodplain, and disaster recovery funding. UGRA described its past and current flood-warning and mitigation efforts, including gauge funding, a new software-based flood prediction project, and consideration of additional retention dams. Kendall and Real counties highlighted successful or needed alerting and evacuation practices, while also stressing the difficulty of funding and maintaining such systems in small rural counties. Members asked detailed questions about the timeline of the flood response, low-water crossings, communications failures, sirens, bridges, and whether regional consolidation or additional infrastructure could improve future preparedness.

Transcript Highlights:

• Instead, our models aim to learn and understand language to predict what comes next and to do so in context
• , allowing our models to statistically predict the next most likely word.
• Developing an advanced language model, like our latest model ChatGPT 4.5, involves two key steps.
• First, teaching the model general intelligence, which is predicting outcomes, reasoning, and problem
• So if I were to prompt, tell me a story about a princess, what the model would do is actually just predict

Summary: The subcommittee met to hear a panel discussion on artificial intelligence and automation in Florida government. Panelists from Worldwide Technology, OpenAI, the James Madison Institute, and the Florida Digital Service described AI as a tool to improve efficiency, constituent services, fraud detection, translation, HR workflows, public safety, and regulatory review, while emphasizing that AI should supplement rather than replace workers. They also stressed the importance of data readiness, workforce training, governance, and having existing technology systems and policies in place before broader deployment. Members asked about public records concerns, vendor vetting, model integrity, and how to balance innovation with privacy and security. Chief Schoonover said agencies are already using or exploring AI in areas such as Medicaid analysis, environmental data, emergency management, tax forms, and child support, and noted that the Digital Service publishes prohibited vendor and application lists. OpenAI discussed its red teaming, safety evaluations, privacy controls, and government products such as ChatGPT Gov, and said it does not train on personal data or target users with ads. The committee also discussed concrete examples of efficiency gains, including reduced call times, faster translation turnaround, and streamlined HR classification work, as well as concerns about job disruption and constituent access. In closing, members raised data center infrastructure and energy needs, with panelists noting that AI depends on substantial compute, power, cooling, and utility capacity. The chair concluded by encouraging continued dialogue and noting that future AI legislation should be targeted to specific problems rather than overly broad.

Transcript Highlights:

• ALL THOSE TYPES OF THINGS AND WE CAN USE AI ARTIFICIAL INTELLIGENCE TO NOT JUST DETECT THE CROP BUT MODEL
• WE USE THESE TOOLS CAN HELP US PREDICT WHAT WORK TO DO WHEN.
• WE COULD ALSO USE HELP IN TAKING OUR AI MODELS AND REFINING THEM, WE HOPE TO FIND GRADUATE STUDENTS THAT
• THAN CHATGPT OR CLAUDE, THEY'RE USED LOOKING AT WORDS AS TOKENS AND USING A NEURAL NETWORK SO THE MODEL
• WATERMELON YIELD AND PREDICTING TOMATO YIELD SO LET'S START WITH WATERMELON YIELD.

Transcript Highlights:

• So we need a lot of data to make those predictions.
• So in all seriousness, what we do is we take those kinds of predictions and we take those kinds of predictions
• I don't want my child's data being used to train the AI models.
• All models are useful. All models are wrong. Some are more useful than others.
• is doing, and what the model is outputting.

Summary: The Senate Committee on Education Pre-K through 12 met to hear a panel discussion on artificial intelligence in K-12 education. Dr. Maya Israel of the University of Florida gave an overview of AI literacy, the benefits of AI for personalized learning and teacher support, and the risks involving data privacy, overreliance, hallucinations, and misuse. She described the Florida K-12 AI Task Force, which published statewide guidance in March and is now providing professional development, case studies, surveys, and resources for teachers and families. She also noted that teacher preparation programs and microcredentials are being developed, but there is not yet a statewide baseline standard for teacher AI training. Superintendents Van Ayers of Hillsborough County and Kevin Hendrick of Pinellas County described district policies and implementation efforts. Hillsborough adopted an AI governance policy, created an implementation guide, and set rules barring generative AI use for students in pre-K through 7th grade while allowing limited, teacher-approved use for older students with district-vetted tools. Pinellas emphasized digital responsibility, parent transparency, data-use agreements, and a governance process for approving tools. Both districts said AI is being used for lesson planning, tutoring, reading-level adjustments, scheduling, and other administrative tasks, but they do not yet have reliable evidence that AI is improving academic performance. They also discussed teacher training, stipends, and certification efforts, including partnerships with universities and vendors such as Microsoft, Google, Magic School, and Gemini. Drew Allen of FSU Inspire offered an industry-focused perspective, arguing that AI is not new but that education must adapt more quickly to industry needs. He described Inspire’s workforce and research mission, its AI and robotics training for teachers, and the need to build teacher comfort and confidence through incentives and applied training. Committee members raised concerns about student dependence on AI, academic integrity, mental health, moral decision-making, data privacy, and whether schools should return to more paper-based testing. The panel generally agreed that AI should be used with strong guardrails, human oversight, and a balance between technology use and traditional learning. No legislation was considered, and the meeting ended with a motion to adjourn, which was adopted without objection.