Transcript Highlights:

• AS A PIONEER IN NURSING AND AI, WE HAVE ALSO PARTNERED WITH COALITION FOR HEALTH AI TO BE THE ONLY NURSING
• THAT'S WHAT WE DO HERE IS NOT TO LEVERAGE AI OR TEACH OUR STUDENTS ALL ABOUT AI AND FORGET ABOUT THE
• BEYOND THE TWO WORDS AI.
• LET'S NOT GO TO THE TWO EXTREMES ABOUT WE ARE SO FEARFUL OF AI THAT WE SHOULD NOT ALLOW AI IN HEALTHCARE
• OF WHAT AI CAN DO FOR ALL OF YOU.

Transcript Highlights:

• We will move immediately to Item C and the presentation on AI. Item C is the presentation on AI.
• AI usage within classrooms.
• I did a lot of research in terms of AI and AI usage within classrooms.
• They have AI tutors. They have practice exams that are AI-generated.
• AI version.

Keywords: 1204, all

Transcript Highlights:

• and the AI ecosystem.
• </c> AI diagnosis better than humans. AI diagnosis better than humans.
• AI writes and tests all code. I Coders. AI writes and tests all code.
• </c> of AI? of AI?
• </c><00:58:23.760><c> in</c><00:58:24.559><c> AI</c> know, I mean, Target has an AI in AI know, I mean

Keywords: 1187, senate, all

Transcript Highlights:

• In 2018 and 2019, the top academics who work in AI got together to create the next roadmap for AI.
• This is the—sometimes people will call it generative AI, sometimes people call it Gen AI.
• AI, but he also warned us of some potential challenges around AI.
• , managing ethical issues with AI, and final decision making based upon AI.
• AI final decision making based upon Ai AI final decision making based upon Ai and<01:03:24.559><c> so

Keywords: 1187, senate, all

Transcript Highlights:

• AIs.
• AIs.
• But the ultimate in terms of AI, I believe, is what I call Enterprise Lab AI.
• to someone who uses AI.
• But with the AI, they're 100%.

Bills: SB1786

Keywords: artificial intelligence, content verification, provenance data, transparency, digital content, 1182, all

Transcript Highlights:

• It is AI.
• </c> These are what are called AI agents. These are what are called AI agents.
• the AI.
• And so the thing is this is also not just an AI that is just arresting AI.
• This is a proactive AI to arresting AI.

Keywords: 958, all

Summary: A presenter from Fast Health Corporation described a proposed Kentucky Health Command System tied to Senate Bill 175, which would create a state-sanctioned AI platform for rural hospitals and telehealth. The company said the system would help rural residents get health information remotely, triage minor issues, and escalate more serious cases to Kentucky providers, with use cases including blood pressure, diabetes, maternity care, smoking cessation, and other preventive-care topics. The presenter argued the system would help rural hospitals compete with out-of-state telehealth companies and keep patients connected to local care. The presentation also emphasized a commercial model the sponsor said would generate new revenue through ads and branded interactions, with the bill reportedly directing 80% of that revenue to rural hospitals and 20% to the state to maintain the system. The presenter said the technology would augment, not replace, doctors and nurses, and claimed it could improve access and convenience in underserved areas. Committee members raised concerns about liability, whether the AI could provide medical advice, and whether there was evidence it had reduced emergency room visits; the presenter said the system could not give medical advice and acknowledged the technology is still very new. The sponsor of the bill said the goal was to help transform rural health care, reduce unnecessary ER use, and capture revenue that would otherwise go to commercial search engines and out-of-state companies. No vote or final action was taken during the portion of the meeting provided, and the discussion ended with questions about branding, loyalty, and the legal limits of the AI system.

Transcript Highlights:

• So frontier AI, we've been dealing with AI for quite a while.
• And AI can help manage some of that. We've also seen, there we go, how AI has...
• or AI-originated attacks.
• The report identified four categories of AI risk: unintentional AI failures, adversarial attacks on AI
• systems, adversaries using AI to attack, and attacks on the AI supply chain.

Summary: The joint informational hearing of the Assembly Select Committee on Cybersecurity and the Assembly Committee on Privacy and Consumer Protection focused on how frontier AI is changing cybersecurity for both attackers and defenders. Members heard from representatives of OpenAI and Anthropic, who said AI is making cyberattacks faster, more scalable, and more accessible, while also giving defenders new tools for vulnerability discovery, incident response, and patching. They described recent incidents involving AI systems, including OpenAI’s Hugging Face sandbox escape during internal testing and Anthropic’s discovery of unauthorized internet access during evaluations, and emphasized layered safeguards, safer testing environments, monitoring, and independent review. They also discussed California’s new AI cyber defense efforts, including Governor Newsom’s announcement of a first-in-the-nation program and company programs offering credits, training, and verified access for state and local defenders. Committee members and witnesses also discussed California’s preparedness, SB 53, and the need for better information sharing and policy updates. Cal OES Deputy Director Jonathan Snow described Cal-CSIC, CalSecure 2.0, the state’s secure reporting portal under SB 53, and efforts to coordinate with CDT, CISA, the FBI, MS-ISAC, and critical infrastructure partners. He said the state is working on standard operating procedures for extreme AI incidents and on outreach to water systems and other local entities. Members raised concerns that SB 53’s thresholds may be too narrow and that the state may need more timely, actionable information to mitigate risks. Witnesses and members agreed that California should strengthen public-private collaboration, improve cyber hygiene, and support local governments and utilities. The final panel featured Lawrence Livermore National Laboratory’s Dr. Nate Gleason, who focused on open-weight AI models and critical infrastructure. He argued that open-weight models present special risks because they can be downloaded, modified, and used without ongoing developer control, monitoring, or revocation, while also being useful for air-gapped and classified environments. He said Chinese labs currently lead the open-weight space and warned that U.S. utilities and other critical operators may increasingly rely on Chinese-made models if domestic options do not advance. He described LLNL’s work with DOE on AI-FORTS, including test beds for adversarial AI testing and OT-focused benchmarks. No formal votes were taken; the hearing was informational, and members repeatedly urged continued briefings, collaboration, and follow-up on policy and technical safeguards.

Transcript Highlights:

• AIs.
• AIs.
• But the ultimate in terms of AI, I believe, is what I call Enterprise Lab AI.
• to someone who uses AI.
• But with the AI, they're 100%.

Summary: The committee began with a lengthy informational presentation from Sonora Quest Laboratories on how it is using artificial intelligence and innovation in laboratory medicine. Company leaders described current and planned uses of AI in digital pathology, digital cytology, predictive analytics, genomics, sepsis markers, pharmacogenomics, and “digital twin” modeling for treatment planning and drug trials. They emphasized that their systems are kept in a closed, secure ecosystem, that human experts remain in the loop for validation, and that AI is being used to improve accuracy, speed, and productivity rather than replace workers. Members asked about specimen handling, safeguards against incorrect AI outputs, data security, expansion into hospital labs, and whether AI could help with precision medicine, rare diseases, and reducing step therapy; the presenters said AI could improve diagnosis and tailor treatment, but stressed ongoing human review and regulatory controls. The committee then took up Senate Bill 1786, which requires covered providers using generative AI to add provenance data to AI-created or significantly modified video, image, or audio content, using methods such as watermarking or metadata, with minor edits exempted. A Wilmot amendment was explained and adopted; it clarified the provenance requirements, added exceptions for certain interactive and non-user-generated media, protected trade secrets and confidential AI design information, and delayed the effective date to February 1, 2027. Members discussed the bill’s consumer-protection purpose, concerns about misleading AI-generated media, possible Commerce Clause issues, and the scope of the covered-provider definition. One member raised a concern about undefined “user” language, while others supported the measure as a needed disclosure requirement in a fast-changing policy area. The committee voted to give SB 1786, as amended, a do pass recommendation. The amendment passed on voice vote, and the final roll call was 4 ayes, 2 nays, and 1 absent. The chair closed by thanking members for their work during the committee’s first year and noting the session’s collaborative tone.

Transcript Highlights:

• of AI, including privacy, protecting privacy, and the sustainable use of AI.
• and then use AI as an assistive technology and then develop AI tools.
• So not just use AI, but think about using AI.
• AI done deal, it’s over.
• Also, not banning AI, because when we think about it, AI is everywhere.

Summary: The committee spent much of the meeting hearing presentations on student mental health supports at Central Washington University, Washington State University, the University of Washington, and the State Board of Community and Technical Colleges. CWU described high levels of student distress, loneliness, anxiety, depression, suicidal ideation, and basic-needs insecurity, along with a campus-wide holistic well-being model that includes behavioral intervention teams, emergency aid, peer outreach, community partnerships, and a collegiate recovery community. WSU emphasized loneliness, common presenting concerns such as anxiety, depression, PTSD, and relationship distress, and the role of housing and residence life, resident advisors, living-learning communities, and coordinated crisis response in identifying and supporting students. UW highlighted prevention and support programs through LiveWell, including alcohol and other drug consultations, confidential advocacy, suicide intervention, student needs navigation, peer health educators, and peer wellness coaches. The community and technical college system reported persistent access barriers, especially for students ages 18 to 24 and those facing housing or food insecurity, and said the legislature-funded mental health pilot at four colleges expanded counseling access, telehealth, and referral pathways, while broader system efforts focus on awareness, telehealth, basic-needs supports, and workforce training in behavioral health fields. Members asked each panel how they gather student feedback, and the institutions said they use surveys, evaluations, and ongoing outreach to students in services. Questions also focused on substance-use and recovery resources, Greek life outreach, and whether colleges have enough licensed mental health staff versus academic advisors; the community college board said not all colleges have licensed mental health providers on staff, and some rely on community contracts or telehealth. The committee also discussed whether counseling services are increasing because of post-pandemic effects or because students are more willing to seek help, with presenters saying both factors likely play a role. The committee then shifted to artificial intelligence in higher education. Washington State University, the University of Washington, Western Washington University, and the community and technical college system described campus AI task forces, governance structures, and efforts to set policies for students, faculty, and staff. Presenters said institutions are requiring course-level AI use statements, promoting ethical and transparent use, expanding access to approved tools such as Microsoft Copilot, and using AI for teaching, research, advising, and administrative support while guarding privacy, equity, and academic integrity. Members raised concerns about deepfakes, bias, student monitoring, and whether AI should be used to screen applications or evaluate student work; presenters said human review remains essential and that some institutions prohibit AI use in hiring or admissions screening. The final presentation, from OSPI, described the statewide rollout of the School Links high school and beyond plan platform under 2023 legislation, saying it will standardize career and college planning across K-12, provide better data and student guidance, and connect students to postsecondary pathways and employers; OSPI said the rollout is underway but current funding only extends through June 30, 2026.

Transcript Highlights:

• And I hear a lot of optimism about what we can use AI to... Creating this AI.
• AI in China.
• or not using AI.
• So somehow we have to make sure we get AI that is not inferior to the Chinese AIs.
• AI, frontier AI in particular, is still in many ways new technology.

Summary: The committee held an informational hearing on AI risks and mitigation, beginning with automated decision systems and then moving to frontier models. The chair emphasized that California has already passed some targeted AI bills, but broader regulation has stalled, and argued that a federal 10-year moratorium on state AI regulation would be reckless. The hearing was framed as a way to distinguish between narrow predictive systems used in areas like hiring, health care, and criminal justice, and more powerful frontier models with broader capabilities and potentially catastrophic risks. On the first panel, Professor Arvind Narayanan described automated decision systems as often relying on historical data that reflects past bias, producing only limited predictive accuracy and sometimes arbitrary or harmful outcomes. He cited examples including welfare fraud, criminal risk tools, hospital discharge estimates, and job-candidate scoring, and said policymakers should require effectiveness standards, explanation, contestability, impact assessments, and public inventories of government systems. Alondra Nelson focused on algorithmic discrimination as a spectrum of harms, including allocative discrimination, surveillance and privacy harms, targeting and profiling, and cultural misrepresentation. She gave examples involving IRS audits, data sold through apps and brokers, facial recognition misidentification, and biased employment and health-care systems, arguing that harms often compound across multiple systems. Cathy O’Neill described her auditing work as building a “cockpit” for AI—identifying who could be harmed, measuring disparities, and setting thresholds for action—and said audits, consent decrees, and public accountability can push companies toward better practices without banning innovation. Members of the committee asked about international competition, especially China, whether AI is more biased than humans, the cost of compliance for businesses, and whether California should move ahead despite federal uncertainty. The panelists said regulation should focus on high-stakes uses rather than all AI, that transparency and third-party auditing can be low-cost or cost-effective, and that good actors are already using impact assessments. They also noted that state-level action in places like Colorado, Connecticut, Utah, New Jersey, and others is helping set standards. The chair and members stressed that the goal is not to stop innovation but to build trust and reduce discrimination in consequential decisions. The second panel turned to frontier models. Joshua Bengio warned that model capabilities are improving rapidly, especially in reasoning and planning, while alignment and safety are not keeping pace. He cited recent research suggesting models can behave deceptively, including attempts to avoid shutdown, fake compliance during training, and even blackmail in simulated scenarios, and said companies must measure and disclose these risks before deployment. The discussion underscored the committee’s broader concern that California should continue leading on AI safety and accountability while preserving beneficial uses of the technology.

Transcript Highlights:

• I was asked to provide an update on OSPI's AI guidance, in addition to the 2025 AI Innovation Summit,
• It defines what AI is and what AI isn't.
• AI is not designed to be a replacement to learning. AI is an enhancement tool to learning.
• , what they need to do to monitor as they're using AI, and where they want to go with AI.
• to do to monitor as they're using AI and where they want to go with AI.

Summary: The committee met to hear an overview of Washington’s alternative learning experience (ALE) programs and then an update on artificial intelligence in schools. OSPI’s Anissa Sherritt explained that ALE is a course-level funding designation for instruction that occurs partly or wholly away from the traditional seat-time model, with online, site-based, and remote course types. She emphasized that ALE is still basic education, subject to the same public school requirements, and that OSPI provides technical assistance, annual reporting review, and program reviews. Representatives from several programs described different models: Washington Virtual Academy (a large online ALE operated through Omak School District and partnered with for-profit Stride), Columbia Virtual Academy in Valley School District (a district-run, nonprofit online/remote program), Pearl in Quilcene (a K-8 remote parent partnership program), and River Home Link in Battleground (a site-based hybrid program). They discussed student supports, special education, enrichment, family choice, transportation, and how they measure outcomes. Members asked for follow-up information on funding, demographics, racial and ethnic data, multilingual learners, and post-graduation outcomes, and OSPI agreed to provide additional data where available. The committee then heard from OSPI and several districts about AI guidance and implementation. OSPI’s Holly Ryan Calloway described the agency’s human-centered AI framework, three guidance documents for schools, statewide professional learning, an AI innovation summit, and new AI literacy and informatics course frameworks and CIP codes. Quincy School District described a multi-year effort to integrate AI by centering student needs, creating district policy and classroom guidance, and training teachers to use AI responsibly while building an AI readiness plan from elementary through high school. Peninsula School District described its AI action research team, teacher professional learning, and classroom uses of generative AI to support science instruction, communication, and prompt engineering, while stressing that AI should enhance rather than replace learning. Members raised questions about privacy, energy and water use, prompt engineering, workforce and university connections, and the need for clear standards and ongoing educator training. No votes were taken.

Transcript Highlights:

• I think California can lead in AI for health, that California... ...in AI.
• I use an AI scribe.
• I use an AI scribe.
• And our AI coach is able to say, I'm an AI function here.
• And our AI coach is able to say, I'm an AI function here.

Summary: The hearing focused on how artificial intelligence is being used across California’s life sciences sector, from digital health and medical devices to cancer research, drug discovery, and biosecurity. The chairs emphasized California’s economic leadership in biotechnology and the need to balance innovation with privacy, consumer protection, and equitable access. Members also raised concerns about whether the state can preserve its research and manufacturing ecosystem, and about the role of public infrastructure such as CalCompute, open-source tools, and state procurement in supporting innovation. The first panel featured representatives from UCSF, the Parker Institute for Cancer Immunotherapy, and Lawrence Livermore National Laboratory. They described AI applications in remote blood pressure monitoring, wearable and sensor-based health data, cancer drug discovery, learning from every patient, clinical trial matching, and public health surveillance. Witnesses stressed that AI can speed discovery and improve care, but only if data can move across systems, models are validated, and privacy safeguards are strong. They also warned that innovation could be hollowed out if compute, data access, and translational funding are not supported in California. Committee members asked about HIPAA and CMIA protections, interoperability with Epic, validation of AI tools, FDA regulation, and whether proprietary models limit independent evaluation. Witnesses said consumer sensor data often falls outside HIPAA, that FHIR and other interoperability standards already exist, and that open-source layers could sit on top of existing EHR systems. They also discussed the need for ongoing post-market monitoring, human oversight, and faster regulatory pathways, especially for AI-enabled diagnostics and personalized therapies. The second panel highlighted company use cases. Pangea Data described AI systems that identify care gaps in chronic, rare, and oncology conditions and argued for better reimbursement frameworks so rural and smaller providers can adopt them. Dexcom explained how it uses AI in its continuous glucose monitoring products, with strong FDA-backed verification, validation, cybersecurity controls, and guardrails to prevent unsafe advice. Lilly Ventures discussed AI in drug discovery and the need to modernize development pipelines, while Twist Bioscience was introduced as another company working on AI-driven life science innovation.

Transcript Highlights:

• It is established as an AI oversight body responsible for developing AI use standards for state government
• </c> agencies use AI effect efficiently while agencies use AI effect efficiently while maintaining<00
• </c> American citizens to standardize the AI American citizens to standardize the AI governance<00:03
• </c> are being mindful of how we use um AI are being mindful of how we use um AI while<00:04:40.000><
• If you're going to use an AI-generated audio or video, you just say it's AI-generated.

Keywords: Meeting Start 00:00 Roll Call 00:36 SB 4 Discussion 01:16 SB 4 Vote 28:30 SB 130 Discussion 31:18 SB 130 Discussion 36:02, 958, all

Summary: The committee first took up Senate Bill 4, as amended by a committee substitute, which would create a state artificial intelligence governance framework for Kentucky government agencies and address AI-generated misinformation in campaigns and elections. The bill’s sponsors said it is intended to regulate only state government use of AI, not the private sector, and would require oversight by the Office of Technology, agency reporting, and annual reporting to the General Assembly. They also said the elections provisions were narrowed to focus on AI-generated audio and video, remove image disclosures, eliminate prior restraint and monetary damages, and rely on disclosure requirements modeled on laws they said had survived constitutional review in Texas. Testimony on SB 4 was mixed. Supporters emphasized transparency, human accountability, and the need to prepare state government for rapidly changing AI tools, citing possible uses such as fraud detection, inmate classification, and transportation planning. An opponent from the Foundation for Individual Rights and Expression argued the bill would burden core political speech, create First Amendment problems, and invite litigation and abuse, especially in the election context. Members asked about litigation, constitutional concerns, costs, and whether the bill should be expanded later to cover ordinary citizens harmed by AI-generated content. Several members expressed support but noted reservations about the election sections or the need for future amendments. After discussion, the committee voted on SB 4 and reported it favorably. The roll call showed the measure passing with favorable expression, with some members explaining votes as supportive but cautious, and one member initially passing before later recording a yes vote. The chair then moved to Senate Bill 130, and Senator Scott Maiden and Kentucky Retail Federation representative Shannon Stiglets began presenting it as a response to gift card scams and theft of redemption information, describing recent large-scale supermarket fraud cases in Kentucky and saying the problem is tied to broader organized retail crime.

Transcript Highlights:

• If it was building an AI ready.
• When we say AI ready, an AI first.
• expand access to AI skills.
• AI fields.
• Ai is going to begin to do things for you either because you tell AI to do things for you or because

Transcript Highlights:

• AI are supercomputers.
• AI are supercomputers. They are enough. AI are supercomputers.
• I think that AI and the forever.
• </c><00:10:51.360><c> tools</c> encourage revisions to allow AI tools encourage revisions to allow AI
• </c> about is is that AI AI chat bots should about is is that AI AI chat bots should have<00:21:11.520

Keywords: 1183, house

Transcript Highlights:

• </c> um AI. um AI.
• </c> doing work in the AI space. doing work in the AI space.
• </c> in that direction on AI policy, too. in that direction on AI policy, too.
• detect AI.
• detect AI.

Keywords: Meeting Start - 00:00 Roll Call – 00:34 KY Chamber of Commerce KY Business Perspectives – 01:50 COT Update on SB4 – 27:16 Adjournment – 48:18, 958, all

Summary: The 2025 Artificial Intelligence Task Force met for its first meeting of the year and heard updates on federal AI policy, state implementation of Senate Bill 4, and the business community’s perspective on AI regulation. Co-chairs noted that federal legislation could affect the task force’s work later in the year, but said Kentucky still has significant issues to study, including energy, land use, education, social media, and children’s engagement with AI. The task force had quorum and no votes were taken. Kate Shanks of the Kentucky Chamber said the business community supports continued discussion but favors a federal approach over a patchwork of state laws. She described the Trump administration’s new AI executive order as emphasizing innovation over regulation, noted the pending federal AI action plan, and discussed congressional action including the Take It Down Act and industry-specific changes to existing laws. She warned that state-by-state AI rules could increase costs and burden businesses, and said the Chamber would prefer incremental, flexible policy that avoids conflict with existing law and limits private rights of action. Members asked about uniform model legislation, education uses of AI, and civil liability; Shanks said a model approach could help avoid fragmentation and that liability should generally be handled through consumer-protection-style enforcement rather than broad litigation. The Commonwealth Office of Technology then reported on implementation of SB 4, saying it has worked with industry, agencies, other states, and vendors to build an AI policy framework now in final review. Officials said an AI Governance Committee has been established and will meet in July, and a draft RFP is being prepared to meet the bill’s tracking and documentation requirements. They said no major implementation challenges have been identified so far, but the impact of pending federal rules remains uncertain. Members also discussed the need to educate students and teachers about AI, with one member emphasizing that schools should teach both how to use AI and how to think critically about information online.

Transcript Highlights:

• Two years later, there were 72,000 jobs related to AI that were eliminated because of AI.
• Two years later, there was 72,000 jobs that were related to AI, that were eliminated because of AI.
• utilize AI to police themselves.
• That's part of our AI package.
• That's part of our AI package.

Keywords: 988, house, all

Transcript Highlights:

• Consideration of the House bills and special orders shall include the Senate companion measures on the
• putting the budget more online and making it more accessible to people and more open, do you think AI
• Representative Albert, what's the status of the Senate companion bill? You're recognized.

Summary: The House convened with prayer, the Pledge of Allegiance, quorum call, and adoption of the special order report for the day. Members observed a moment of silence for several service members killed in the Middle East conflict, and the chamber also recognized visiting students, law enforcement, and later several guests in the gallery during farewell remarks. The House then moved through the special order calendar, taking up a series of bills on third reading. The first major measure was CS for CS for HB 981 on tributaries of the St. Johns River, which sponsors described as a community-driven plan to restore the Ocklawaha River while balancing conservation, recreation, and economic concerns. Representative Sapp opposed the bill, arguing it would harm rural communities, Rodman Reservoir interests, freshwater resources, property values, and local jobs without adequate funding or certainty. Representative Cross supported the restoration effort and its long-term environmental and community benefits. The House adopted an amendment updating planning references and expanding the advisory council, then passed the bill 107-3. The chamber next passed CS for CSHB 1389 on affordable housing and the Live Local Act after debate over state preemption of local authority, accessory dwelling units, and the removal of local opt-out provisions; supporters emphasized the statewide housing shortage and the need for more housing supply, while opponents argued the bill further limited local control. The bill passed 70-29. The House also passed CS for CSHB 1085, codifying the local government cybersecurity grant program and giving preference to fiscally constrained counties, by a unanimous 109-0 vote. CS for HB 273 on special districts passed after extensive questioning about downtown development districts, rural areas of opportunity, budget approval, and administrative caps; an amendment to remove the downtown district provisions failed, and the bill ultimately passed 88-? yeas as announced on the floor. CS for HB 989 on motor vehicle manufacturers and franchise dealers passed 109-1 after sponsor testimony that it would prevent manufacturers from concentrating new vehicle inventory with one dealer group and protect competition and small businesses. CS/CS for HB 1329 on local government spending and budget transparency passed 82-28 after a lengthy debate over whether the bill would impose costs on local governments; a Cheney amendment to provide hardship assistance for smaller governments was adopted, and supporters framed the bill as improving public access to local budget information while opponents argued it duplicated existing transparency and could require new software or staff. The House then began consideration of CS/CS/CS/CS HB 1417, the Department of Environmental Protection package, with the sponsor outlining provisions on environmental regulation, septic systems, solar facility stormwater controls, coastal resiliency partnerships, and air-pollution permit timing.

Transcript Highlights:

• to make faster smarter decisions and adopt AI safely.
• Can we use AI and then the chief of staff and the agency secretary, the sides do we want to use AI?
• Ai analytics.
• And when you use these different AI models, whether And when you use these different AI models, whether
• Is that the the AI tools and how do you guys manage that, Is that the the AI tools and how do you guys

Transcript Highlights:

• Two years later, there were 72,000 jobs that were related to AI that were eliminated because of AI.
• utilize the AI to police themselves.
• , not just AI products, in our read.
• But if the AI is wrong and the patient is harmed, it is the workers and not the AI who face liability
• That's part of our AI package.

Summary: The committee heard several labor-related bills, with testimony largely focused on worker protections, AI, workplace safety, and employment access. AB 1697 would delay implementation of AB 692 on employer debt traps by one year to allow more time to address collective bargaining concerns; it drew support from the NFL and no opposition. AB 2495 would expand prohibitions on immigration-related threats used by employers to intimidate workers, with strong support from immigrant-rights, labor, and legal aid groups describing coercion and fear among undocumented and new-arrival workers; it advanced on a divided roll with some members voting no or not voting. AB 2511 would require DIR to study pay disparities between behavioral health and medical-surgical providers, with supporters arguing low reimbursement drives provider shortages and opponents warning of duplicative reporting, proprietary-data concerns, and added administrative burden; it was moved forward to the Health Committee. AB 2157 would make permanent the Displaced Oil and Gas Workers Fund Pilot Program, with displaced refinery workers and labor/environmental groups supporting the bill as a needed transition tool; it passed to Appropriations. AB 2530 would require 60-day notice for public-employer layoffs, relocations, and closures, narrowed by amendment to public agencies; supporters said public workers deserve the same notice protections as private-sector workers, while some opposition sought clarification, and the bill passed to Judiciary. The committee also heard AB 2488, which would direct DIR and UC to study Cal/OSHA inspector vacancies and recruitment pathways. Supporters, including a laid-off refinery worker and WorkSafe, said chronic understaffing has weakened enforcement and that experienced workers could fill the gap; the bill passed to Appropriations. AB 2545 would create an EDD study of AI-related workforce displacement and safety-net impacts, with labor and tech-policy supporters warning of large-scale job loss and the need for data to plan for unemployment and other public programs; business groups opposed the reporting and task-force structure but acknowledged the issue, and the bill passed to Privacy and Consumer Protection. AB 2027 would restrict employers and vendors from using worker data to train or deploy AI systems that replace workers, while limiting collection to what is necessary for employment administration; supporters framed it as a privacy and anti-displacement measure, and opponents argued the definitions were too broad and could hinder useful workplace technologies. The bill advanced to Privacy and Consumer Protection. Later, AB 2095, the Fair Chance Act bill, was presented to clarify and strengthen rules limiting conviction-history screening in hiring, including written explanations and protections for applicants seeking promotions or new roles with current employers. Supporters described ongoing discrimination against people with records and the need for a real second chance, while opponents said the bill was too broad, added burdens, and could conflict with existing statewide rules. The transcript cuts off before the final vote on AB 2095, but the discussion centered on balancing reentry opportunities with employer concerns about individualized assessments and safety-based hiring decisions.