Recruiting & HR: California puts a person behind every AI firing

Recruiting and HR now have legal deadlines for human judgment in AI-driven personnel decisions. On 30 September 2026 the governor of California, the largest US state economy, signed a package of AI bills, three aimed at the workplace: from 1 January 2027 employers may not use AI to read workers' emotions, and layoffs caused by automation must say so; from 1 July 2027 no worker may be disciplined or fired on software's word alone. Meanwhile recruiting agents are arriving inside tools teams already pay for, before anyone has shown they speed up recruiting.

California sets dates for the human check

Gavin Newsom signed the package on his last major signing day before he leaves office in January (governor's office and the Associated Press, 30 September 2026). Three of the bills matter to anyone who employs people in the state.

The first, SB 947, called the "No Robo Bosses Act", bars an employer from relying only on an automated system to discipline or fire someone. The person reviewing the case must collect their own evidence before acting. A worker disciplined or let go through a process that relied mainly on such a system must get a written notice saying so, a description of the data the system used, and the name of a person who can explain the decision. It covers independent contractors and takes effect on 1 July 2027. Newsom vetoed an earlier version in October 2025. The law covers discipline and firing; hiring decisions are outside it.

The second, AB 1883, was awaiting signature when Era Haus last looked at recruiting and hiring. It is now law. From 1 January 2027, California employers may not use monitoring tools that infer a worker's emotions from face, voice or tone, or that collect brain-activity data, with fines of up to $500 per violation (Bloomberg Law, 30 September 2026).

The third, SB 951, changes California's mass-layoff notice law, which requires advance warning before large layoffs. From 1 January 2027, a notice for a layoff caused wholly or largely by AI or automation must say so in its heading and describe the tasks being automated, according to the law firms Littler and Ogletree. It applies only to employers the notice law already covers, generally those with 75 or more staff, so most small firms are outside it. It also asks the state's employment department to report on employers that stop hiring for jobs because of the technology, which is the clause to watch.

The recruiting agent arrives inside tools you already pay for

An AI agent here means software that runs a chain of recruiting tasks on its own: finding candidates, messaging them, booking the first conversation. Two moves in September put agents into daily recruiting work with no new purchase decision.

LinkedIn announced the second version of its Hiring Assistant at its Talent Connect conference in New York at the end of September (HR Brew and ERE, a recruiting trade site, late September 2026). It learns how each recruiter has hired before, flags whether a candidate has verified their identity and seems open to a move, and works with outside applicant tracking systems, the software where a company keeps its job ads and applications. It reaches English-language customers in November at no extra cost. LinkedIn told ERE that more than 20,000 companies already use its recruiting agents.

Adecco, the Swiss staffing group, said it is rolling out Salesforce's agent software to about 27,000 of its staff in more than 40 countries, after pilots in the UK and France (company release, 15 September 2026). Its recruiters use it to shortlist candidates and to start pre-screening and onboarding steps. When one of the world's largest staffing firms puts agents into front-line work across Europe and beyond, smaller agencies are likely to meet the same tools through their own software vendors.

Faster hiring has not shown up yet

The promise these tools sell is speed, and the broad figures do not show it yet. ManpowerGroup, the US staffing group, asked more than 40,000 employers worldwide in September how quickly they fill roles compared with 2025: 28% said faster, 41% saw no change and 30% said slower. The survey does not split answers by AI use, so it cannot show the tools failing. It does show that filling roles faster is still a minority experience.

In Argentina, small firms are further behind. A survey of managers at small and mid-size Argentine companies, run by the research firm Trendsity for Deel, a payroll company, found 39% use AI in at least one HR process, falling to 20% among the smallest firms (BAE Negocios, 1 September 2026).

What it means for you

If you employ anyone in California, staff or contractors, "a human made the call" becomes something you must prove in writing, with dates attached. Outside California, the same agents are coming to you through LinkedIn and your hiring software, at no extra cost. An agent that learns from your past hires also learns your past habits, including the ones that narrowed who you interviewed. And for most employers the time saving is still a promise, so do not let it be the reason you change how you recruit.

What to do about it

First, if you have people in California, list every tool that scores attendance, performance or tone of voice. Emotion or "engagement" scoring of staff now has a legal end date: switch it off before 1 January 2027. Before July 2027, write down the step where a manager checks the evidence for themselves before any discipline, and draft the notice a worker would receive.

Second, if you use LinkedIn's recruiter tools in English, when the new assistant arrives in November, look at what it says it learned from your past hires before you let it shortlist. If its picks look like your last ten hires, rewrite the brief you gave it.

Watch, don't act yet: whether California extends the same duty to hiring decisions, and the state report on employers that stop hiring for jobs because of the technology. Either would bring hiring decisions under the same written-proof duty.