Week in review #42
California enacts broad new AI employment laws.
New California AI employment laws
The state of California has enacted several new laws regulating how employers can use AI in the workplace. Effective January 1, 2027, one law prohibits the use of AI-powered surveillance tools that infer an employee’s emotional state or collect “neural data”. Another, effective July 1, 2027, prevents employers from relying solely on automated systems for disciplinary or termination decisions, requiring human corroboration. A third law adds AI-related disclosure obligations to the state’s mass layoff notification requirements, similar to a new rule in Connecticut.
theguardian.combakerdonelson.comduanemorris.com
Connecticut mandates AI layoff disclosures
A new law in Connecticut requires employers to disclose the role of AI in mass layoffs. Effective October 1, 2026, the Artificial Intelligence Responsibility and Transparency Act mandates that any employer executing a mass layoff must provide written notice to the Connecticut Department of Labor stating whether AI or other technology was a contributing factor. The rule is part of a growing trend of state-level legislation aimed at increasing transparency and accountability for automated decision-making in employment.
secureprivacy.aiebglaw.comhinshawlaw.com
Meta adjusts AI workforce strategy
After significant investment in AI and related job cuts, Meta has reportedly scaled back its plans to replace more positions with AI agents. The adjustment follows reports of employees spending increasing amounts of time addressing technical problems and security concerns linked to the new systems. The company is also said to be expanding parts of its management structure, reversing some earlier reductions and indicating that integrating AI does not always lead to expected efficiency gains without continued human oversight.
Meta releases Muse Spark Open model
Meta has launched Muse Spark Open, an 816 billion-parameter open-weight AI model it claims is the most capable of its kind from the US. The model uses a mixture-of-experts architecture, activating 42 billion parameters per token, and supports a context of up to 1 million tokens. It accepts text, image, video, and audio inputs to produce text output. Meta stated the model is optimized to run on Apple’s advanced M5 Ultra chips, and its license permits modification and redistribution, though a separate contract is required for companies with annual revenue over $20 million.
theguardian.comhuggingface.comashable.com
50% of companies see value from AI
Half of all companies are now generating tangible value from their AI initiatives, according to a new index from Boston Consulting Group. The survey of over 1,300 senior leaders found that enterprise spending on AI has more than tripled in the past year, with many firms moving from experimentation to production. The report also highlights that companies considered mature in their AI adoption achieve 2.4 times more top-line growth than less mature peers, demonstrating a clear link between scaled implementation and commercial returns.
bcg.comzdnet.comentrepreneur.com
AI’s mixed impact on European jobs
A new report from the European Commission has identified AI as a source of both opportunity and risk for the European labor market. The analysis found that while 75% of EU firms using AI report productivity gains, only 12% see increased revenue. The report also notes a significant adoption gap, with just 20% of EU firms using AI. Furthermore, only 32% of European workers used AI in early 2026, compared to 43% in the US, highlighting a need for strategic adoption to manage potential job shifts and develop new skills.