News-May-29-2026

May 29, 2026 — Illinois lawmakers sent SB 315, the Artificial Intelligence Safety Measures Act, toward the governor after the Senate recorded the bill as having passed both houses; Shift launched a New York City home-cleaning service that records homes to collect AI robot-training data; CNBC reported that the European Commission wants intensified U.S. talks over advanced AI models with cyber capabilities after Anthropic's Mythos rollout raised government-access and safety concerns; and Transparency Coalition's weekly update showed California AI bills clearing the crossover deadline while Colorado and Missouri chatbot and AI-therapy bills moved toward final action.

Contents

1. Illinois passes frontier AI safety audit bill
2. Shift offers free home cleaning in exchange for robot-training video
3. EU and U.S. discuss Mythos cyber model access
4. State AI bill roundup highlights California, Colorado, Missouri, and New York

---

Illinois passes frontier AI safety audit bill

The Illinois General Assembly's SB 315 page listed the bill's last action on May 29, 2026 as "Senate: Passed Both Houses." The measure creates the Artificial Intelligence Safety Measures Act and applies to large frontier AI developers. NBC News reported that the Illinois House passed the bill on Wednesday and that it would set a new standard for regulating leading AI companies if Governor JB Pritzker signs it.

SB 315 requires covered developers to create, implement, publish, and annually update a frontier AI framework addressing catastrophic-risk assessment, mitigations, governance, cybersecurity, third-party evaluations, and internal-use risks. The bill also requires transparency reports before deployment of new or substantially modified frontier models and mandates annual independent third-party audits. NBC News reported that mandatory annual independent third-party audits would be a first for U.S. AI legislation.

The bill authorizes the Illinois Emergency Management Agency and Office of Homeland Security, in consultation with the Attorney General, to administer reporting systems, issue rules, and compile annual reports. The bill also establishes reporting requirements for critical safety incidents and internal-use catastrophic-risk assessments, creates whistleblower protections for covered employees, prohibits retaliation, and establishes civil penalties for violations.

---

Shift offers free home cleaning in exchange for robot-training video

Ars Technica reported that German startup MicroAGI's Shift app is offering New York City residents free home cleaning by sending cleaners wearing cameras to record household tasks for AI-driven robot training. Semafor reported that Shift launched publicly on May 28, 2026 and that MicroAGI already oversees data-collection work in several countries, anonymizes the data, and sells it to AI labs.

Ars Technica reported that the booking flow asks customers for a phone number, email address, home address, and access instructions before a roughly two-hour appointment. Ars Technica also reported that Shift's FAQ says names, faces, and other personal information are automatically anonymized and that sensitive details are blurred before video is used. The launch is relevant to AI privacy law because it links domestic-space video capture, automated anonymization claims, and the commercial sale or use of household data for embodied-AI training.

---

EU and U.S. discuss Mythos cyber model access

CNBC reported on May 29, 2026 that a European Commission spokesperson said the EU is seeking to "intensify" discussions with the U.S. administration on the most advanced AI models, including models with cyber capabilities. The report identified Anthropic's Mythos model as the immediate context for the talks and said Anthropic had not granted the EU AI Office or government organizations outside the United States, except the U.K. AI Security Institute, preview access to review the model.

Anthropic's Project Glasswing page describes Claude Mythos 2 Preview as an unreleased frontier model that can surpass all but the most skilled humans at finding and exploiting software vulnerabilities, and says Anthropic is using the model defensively with selected launch partners and more than forty additional organizations that build or maintain critical software infrastructure. CNBC reported that Anthropic said models with Mythos-level capabilities require strong cyber safeguards before general release and that the company expects to bring "Mythos-class" models to customers in the coming weeks.

The development is legally significant because it links frontier-model access, cybersecurity safeguards, and cross-border government oversight under the EU AI Act implementation framework and U.S. AI safety policy.

---

State AI bill roundup highlights California, Colorado, Missouri, and New York

Transparency Coalition reported on May 29, 2026 that nearly all of California's thirty remaining AI-related bills had been approved by their chamber of origin before the May 29 crossover deadline. The roundup said California legislators then faced a four-week period to move bills through the opposite chamber before the scheduled July 2 summer adjournment.

The same update reported that Colorado Governor Jared Polis was expected to sign HB 1263, a conversational-AI safety bill, and HB 1195, a bill restricting AI use in psychotherapy services. The Colorado General Assembly's HB 1263 page listed the measure as passed and sent to the governor on May 28, 2026. The Colorado General Assembly's HB 1195 page listed the psychotherapy AI restrictions bill as passed.

Transparency Coalition also reported that Missouri lawmakers sent SB 1019 to Governor Mike Kehoe after final passage on May 15, and described the measure as including penalties for advertising an AI chatbot as capable of providing therapy, a mental health diagnosis, or acting as a mental health professional. The update separately noted that New York Senator Kristin Gonzalez's S 9051, a kids chatbot safety bill, had been reported from the Senate Internet and Technology Committee, approved by the Senate Finance Committee, advanced to third reading, and amended on third reading in May.

The cluster of developments is significant because multiple states are moving beyond general AI-disclosure proposals toward sector-specific chatbot, minors, mental-health, and frontier-model governance measures.