|
Issue 18
|
|
The weekly brief California is about to make it illegal for an algorithm to fire you alone.This issue is about the rules, not the job counts: what the first U.S. law on AI discipline would actually give a worker, which states already got there, and how to find out whether software is scoring you right now. September 21, 2026 · 8 min read
|
Sourced from California and Connecticut bill text, employment-law analyses, Resume Builder, and MultiState. | No hype Statutes, vote counts, and dates, linked and attributed. | Why now Newsom’s sign-or-veto deadline lands at the end of this month. |
|
If you read nothing else California’s legislature has passed SB 947, the No Robo Bosses Act. It bars employers from firing or disciplining anyone based solely on an automated system, requires a human to review and corroborate the decision, and gives the worker written notice, a named human reviewer, and a copy of the data used. Newsom has until September 30 to sign. If he does, it becomes the first law of its kind in the country and takes effect July 1, 2027. Connecticut and Illinois already have narrower versions on the books. |
|
|
Nova’s signal The law doesn’t ban the algorithm. It bans the shrug. Once a human has to sign the decision, the decision has a name, and a name can be asked why. Smaller than a ban, bigger than a disclosure. Nova’s call. Find out this week whether any software scores your work. Ask HR in writing. The answer is now a document you may need. Last week we followed where AI moved the jobs. This week we follow who has to answer for a decision about yours. Story 01 walks through what the bill requires, 02 covers why Newsom vetoed the last version and what changed, 03 maps the states already ahead of California, 04 shows how common AI-made personnel decisions already are, and 05 covers the surveillance bill on the same desk and the consultation closing in the UK on the same day.
|
|
This week 5 developments worth your attention.
|
|
01 / Workplace Law The No Robo Bosses Act: what it would actually give youSB 947 cleared the Assembly 53–14 on August 30 and the Senate 28–10 on August 31, and was enrolled to the governor on September 4. The core rule is short: an employer may not rely solely on an automated decision system to discipline or fire a worker, and where such a system assists the decision, a human must review it and independently corroborate it before acting. The bill defines an automated decision system as any computational process derived from machine learning, statistical modeling, data analytics, or AI that replaces human discretion in issuing scores, recommendations, or decisions with a significant effect on workers. Around that rule sit the mechanics that matter to an employee: timely, plain-language written notice after the system is used, naming the system, the data it relied on, and a contact for the human reviewer, with an anti-retaliation clause; the right to request your own system-related data once every 12 months; the right to review by a human who was not involved in the original decision; and enforcement through the Labor Commissioner, public prosecutors, and a private right of action, with a $500 civil penalty per violation plus possible punitive damages and attorney’s fees. If signed, it becomes operative July 1, 2027. Why it matters. Today, contesting an automated write-up means proving the software was wrong. Under SB 947 the burden flips: the employer has to show you the human who checked, and what they checked against. The signal. Crowell & Moring’s read of the enrolled text is that it also spells out uses an employer may not put these systems to at all, naming predictive behavior analysis and compensation decisions. That list is where the next round of litigation would start. Crowell & Moring, on SB 947’s provisions →
|
|
02 / Politics Newsom vetoed this bill last year. Here is what changed.The 2025 version, SB 7, passed both chambers and died on October 13, 2025, when Governor Newsom vetoed it. His message called the restrictions “overly broad,” said the notice requirements were “unfocused” because they reached “even the most innocuous tools” like scheduling and workflow software, and worried that the bill could stop employers using systems that factor in customer ratings to reward strong performers. Senator Jerry McNerney reintroduced the measure as SB 947 on February 2, 2026, rewritten around those objections: notice is now post-use rather than a blanket pre-use disclosure, and the prohibited uses are enumerated instead of implied. The California Chamber of Commerce, which fought SB 7, has not endorsed the new bill, and its policy advocate warned in committee that language covering systems that merely “assist” a decision could sweep in tools that only inform it. But the Chamber has reportedly dropped SB 947 from its list of “cost driver” bills, the ones it says will damage the state’s economy. McNerney’s office says that if enacted it would be the first law of its kind in the nation. Why it matters. Even the veto told you where the line is. Nobody in Sacramento, on either side, is defending fully automated firing. The argument is about how much a tool can help before a human has to sign. The signal. Watch the Chamber’s list, not its statements. A bill coming off the “cost driver” list means business opposition has moved from the principle to the wording, which is usually the last stop before a signature. Fisher Phillips, on the SB 7 veto and what it asked for →
|
|
03 / State Laws California is not first. Connecticut, Illinois, and New York City already moved.Connecticut’s Governor Lamont signed SB 5 in June as Public Act 26-15, the Connecticut Artificial Intelligence Responsibility and Transparency Act. Its employment pieces phase in: on October 1, 2026, anti-discrimination amendments, a developer-and-deployer framework, and a new WARN Act disclosure that requires employers to say when AI is a reason for a mass layoff; on October 1, 2027, pre-decision notice and disclosure duties when an automated tool makes or is a substantial factor in an employment decision. Illinois has been live since January 1, 2026: HB 3773 requires employers to notify employees and applicants when AI is used in hiring, promotion, discipline, or discharge, and prohibits AI that has a discriminatory effect, including zip codes used as proxies. New York City’s Local Law 144 has required bias audits of hiring and promotion tools since 2023. Colorado went the other way in May, repealing and replacing its landmark AI Act with a narrower disclosure law, pushing the start to January 1, 2027, and opening draft rules for comment through October 26. MultiState counts 155 bills on AI in the workplace across state legislatures this session. Why it matters. Your rights on this depend on your zip code. A worker in Chicago has had a right to notice since January. A worker in Texas has none. That gap is the thing to know before you accept a role. The signal. Connecticut’s WARN-style disclosure is the first time an employer will have to state on the record that AI cut the jobs. Given last issue’s finding that most “AI layoffs” are relabeled financial cuts, that one line may do more than any ban. MultiState, on 2026 state AI employment laws →
|
|
04 / Management 6 in 10 managers already use AI to decide raises, promotions, and firingsThe practice these bills regulate is already ordinary. In a Resume Builder survey of 1,342 U.S. managers, published in 2025, 6 in 10 said they use AI tools to make decisions about their direct reports. Among those managers, 78% had used it on raises, 77% on promotions, 66% on layoffs, and 64% on terminations. About one in five said they let AI make the decision without human input all the time or often, and another 24% said they sometimes do. Two-thirds had received no training on managing people with AI, and ChatGPT was the most common tool, followed by Copilot and Gemini. The public is less comfortable than the managers: a Quinnipiac poll in March 2026 found 15% of American adults would take a job where their direct supervisor was an AI program. Why it matters. The law is chasing a habit, not a product. Most of this happens in a chat window, not in an HR system with a vendor name, and that is the hardest kind of automated decision to notice or prove. The signal. SB 947 covers systems that replace human discretion. A manager who pastes your review into a chatbot, asks whether to fire you, and does what it says is exactly the case the human-review rule was written for, and the one with no audit trail. The Hill, on the Resume Builder manager survey →
|
|
05 / Surveillance The other bill on Newsom’s desk bans reading your face and your brainAB 1883 passed its final Assembly vote 56–8 on August 30 and sits beside SB 947 with the same September 30 deadline. It prohibits employers from using workplace surveillance tools that rely on AI to recognize, infer, or predict an employee’s emotional state, or to collect “neural data,” defined as information generated by measuring activity of the central or peripheral nervous system. It does not ban monitoring in general, and safety uses remain allowed; the penalty is up to $500 per violation. The wider trend it responds to is well documented: Gartner found 70% of large U.S. employers now use some form of monitoring software, up from 30% in 2019, the market for these tools is expected to pass $4 billion this year, and the studies Knowable Magazine reviewed found no correlation between how closely workers are monitored and how they perform. The UK is asking a similar question. The Department for Business and Trade’s consultation on workplace monitoring technologies, which covers keystroke logging, screen capture, activity scoring, and algorithmic management, closes at 11:59pm on September 30 and is open to anyone, not only employers and unions. Why it matters. Productivity scores are the inputs to the systems SB 947 governs. One bill is about what can be measured, the other about what can be done with the measurement. Read them as a pair. The signal. Same date, two countries, one theme: consent before deployment rather than remedies after. The UK is explicitly asking whether employers should have to consult staff before switching monitoring on at all. Mondaq, on AB 1883 →
|
|
Signal & chatter A few more things worth knowing.The EU pushed its workplace-AI rules to December 2027. The Digital Omnibus on AI was published in the Official Journal on July 24 and took effect July 27, days before the AI Act’s high-risk obligations, which include employment systems, were due on August 2. Those obligations now apply from December 2, 2027. Transparency duties were not delayed. Explore → | New York City’s AI hiring law was audited, and enforcement failed the audit. The State Comptroller’s December 2025 review called the city’s enforcement of Local Law 144 “ineffective,” citing complaint handling and inaccurate compliance reviews. Employment lawyers now expect a stricter phase, with penalties of up to $1,500 per violation per day. Explore → |
|
|
From the community This week’s color comes from Blind and a widely shared personal account rather than a single Reddit thread. Both are reported experience, not verified cases, and worth reading as what workers say it feels like rather than what the systems provably did. A Blind thread titled “PIP is finally automated” describes performance plans arriving from a dashboard before any manager conversation. Separately, an account circulating since the 2025 layoffs describes a 25-year Microsoft engineer marked “low impact” by a ranking system and let go on his birthday, while colleagues describe unlogged mentoring and firefighting. The common thread is not a robot pressing a button. It is a score nobody can explain arriving before the conversation, which is precisely the “solely” question SB 947 turns on. Blind, “PIP is finally automated” (anonymous, reported experience) →
|
|
FOWL prediction 18 By the end of 2027, at least five states will require a named human reviewer before an automated system can discipline or fire a worker.Connecticut, Illinois, Colorado, and New York City have each landed on the same two ingredients, notice and a human in the loop, from different starting points. States introduced 155 workplace-AI bills this session. If California signs, the model text exists, and model text spreads the way the CCPA did: fast, and mostly copied. FOWL AI · September 21, 2026 · We’ll score this in September 2027.
|
|
Put it to work 3 things to do this week.| 01 | Ask, in writing, whether any automated system scores or ranks your work and feeds discipline, pay, or promotion decisions. In Illinois you are owed notice already; in Connecticut from October 2027; everywhere else, the answer is still useful and the email is your record. | | 02 | Keep your own copy. Screenshot your metrics dashboard, save the text of every review, and note any “productivity” score you are shown. SB 947’s data request would be once per 12 months; your own file is faster and exists today. | | 03 | If you are in California or the UK, the window closes September 30: a short note to the governor’s office, or a response to the Department for Business and Trade consultation, counts. Colorado’s draft rules take comments until October 26. |
|
Career radar AI Governance AnalystThe person who inventories every automated system a company points at its employees and customers, documents what data each one uses, runs the notice and human-review process these laws require, and keeps the audit trail a regulator or plaintiff would ask for. It sits between HR, legal, privacy, and data teams, and it is growing faster than almost any specialism LinkedIn tracks: AI governance demand is up 150% year over year in its 2026 Skills on the Rise report, with roughly 2,000 U.S. postings since January and a median across governance roles of about $169,000. A place to start. Most people arrive from HR operations, compliance, privacy, or data analysis rather than law school. The IAPP’s AI Governance Professional certification is the credential employers screen for, and one certification correlates with about 13% higher pay. Then read one statute end to end; Connecticut’s Public Act 26-15 is the most complete template in the country right now. Explore AI careers → |
|
|
The opportunity board Know where to look.If you want to earn from your expertise while the rules settle, the AI training and evaluation market is still hiring across a wide band: generalist annotation around $15–25/hr, coding and RLHF work $50–65/hr, and specialist review in medicine, law, and finance reaching $250–1,000+/hr on platforms like Mercor and Surge AI. See the full FOWL AI platform breakdown → Governance, risk, and compliance roles with “AI” in the title are appearing at professional services firms, insurers, and large employers first. Our board is refreshed weekly and every link is checked before it ships. Browse this week’s AI jobs board →
|
|
For a year the AI-and-work debate has been about how many jobs. This week’s is about a smaller, older question: who has to answer for a decision. The bills on Newsom’s desk do not slow the software down. They put a name back on the outcome. That kind of change outlasts any one model release, and it is the kind you can actually use. Until next Monday,
TYB
Data scientist & founder, FOWL AI
|
|
One question for you Has software ever scored your work in a way you couldn’t see or contest?Hit reply. I read every response, and it helps shape what we cover next.
|
|
Atlanta · Vibe Code Saturdays Read about the future.
Then build it.A free, hands-on Saturday to build a website, newsletter, portfolio, or useful tool with AI. Beginners welcome. Bring your laptop. First date and venue coming soon. Apply and we’ll follow up with the details.
|
Know someone who’d find this useful?Share this briefing → Forwarded this email? Get the free weekly briefing.
|
|
|
|