Skip to main content
AI Jobs·Report

ANALYSISReported by AI Jobs Report

Fear, boycott, and a hiring reality: safety jobs still lag

Commentators warn, mock, and organize boycotts, but employers list 4,685 AI vacancies across 178 firms, with just 40 in safety, supporting diffusion claims but not a slowdown.

Read the original at AI Jobs Report
8 min read12 viewsBy AI Jobs Report

Commentators warn, mock, and organize boycotts, but employers list 4,685 AI vacancies across 178 firms, with just 40 in safety, supporting diffusion claims but not a slowdown.

A louder panic meets a busy job market

Across the last two days, several prominent voices amplified a sharp turn in the AI conversation toward danger, boycott, and bans. Zvi Mowshowitz writes that "There was already a preference cascade happening where people finally were admitting that they thought AI might well kill everyone." He adds that a resignation "turned that cascade into an avalanche" and that "Senator Sanders and Representative Casar introduced an outright ban on superintelligence and I have to remind myself that happened this week."

Gary Marcus urges direct action, arguing "The time for a public boycott has come." Nathan Lambert captures the media dynamics behind the shift: "fear sells. Fear is the simplest story, the one people cannot look away from."

Set against that, our hiring tracker shows no sign of a pause. We count 4,685 open AI roles across 178 employers, taken from their own job feeds. In the past 30 days, employers opened 1,065 roles in modeling and engineering and 529 in data. Accenture alone lists 1,026 open roles, with 794 opened in 30 days. OpenAI has 162 open, with 84 opened in 30 days, and Anthropic lists 127 open, with 57 opened in 30 days. Whatever happens in discourse, employers are still staffing AI teams.

Our headline jobs-created indicator is counted from job listings and calibrated to the World Economic Forum’s Future of Jobs Report 2025 (11M AI roles created, 9M displaced, by 2030). We do not see boycott effects in the postings yet.

Are average people feeling it? Employers are building for them, not with them

Nathan Lambert argues that despite big rhetoric, the everyday experience remains marginal: "The touch-points that average people have to AI products today are fringe, marginally beneficial, or even just very confusing to them... These are very small benefits." That reads as compatible with our hiring footprint. The largest employers in our data are services and regulated-industry incumbents rather than consumer startups:

  • Accenture: 1,026 open, 794 opened in 30 days
  • PwC: 128 open, 96 opened in 30 days
  • Capital One: 190 open, 121 opened in 30 days
  • Amgen: 159 open, 41 opened in 30 days

That profile suggests enterprises busy with integration work, internal tooling, and domain applications whose benefits will reach the public indirectly and with a lag. Ethan Mollick puts a time horizon on that diffusion, writing: "Independent of policy debates about existential risk, it is worth noting that even if AI development stopped today with the models we have right now, we would still have at least a decade of roiling change throughout much of work, education, and social life as harnesses improve & AI diffuses further." Our figures do not confirm the length of that timeline, but the breadth of hiring across consulting, finance, biotech, autonomy, and cloud tooling is consistent with continued diffusion rather than a pullback.

Safety and governance: rising concern, small hiring footprint

Gary Marcus argues that the labs and government have thin plans for risk mitigation: "They have no serious plan for mitigating the risks of generative AI. The government, too tight with industry, doesn’t either." He doubles down on transparency and monitoring, warning that "Opacity also becomes a potential tool for authoritarianism" and "I am worried, for example, that the criteria may pay insufficient attention to the importance of monitorability, one of the few tools we have to keep an eye on what AI agents are doing."

Zvi Mowshowitz echoes the measurement problem from a different angle. On OpenAI’s latest model, he quotes the claim that "OpenAI claims that Astra is ‘the most intelligent and most aligned [available] model’ in the world" while also highlighting challenges with "severe problems with monitorability." He notes, too, the regulatory process as reported by OpenAI’s leadership: "Oh no. OpenAI President Greg Brockman confirmed that they did the ‘standard testing process together with the government’ and the government did not ask for any changes."

Do employers’ job postings suggest a shift toward safety and evaluation? Our tracker shows 40 open roles in evaluation and safety, with 17 opened and 5 closed in the last 30 days, across 19 employers. That is a small share next to modeling and engineering at 1,912 open roles across 144 employers and data roles at 846 open across 125 employers. Our data cannot speak to internal reallocations or the depth of safety work inside other postings, but the explicit hiring labeled as evaluation and safety remains limited.

On balance, the concern about a monitoring and safety deficit in public commentary is not contradicted by our postings data. It shows employers emphasizing applied engineering over labeled safety hiring.

Builders keep building: Astra in the wild, and AI-accelerated security

Even as the discourse tilts to danger, practitioners report strong results from new models. Sebastian Raschka writes: "Astra is the best model I’ve used so far, and it’s disproportionately good at 3D rendering and animation tasks (relative to other models)." Simon Willison shows one such workflow in practice: "It churned away for 17m51s and built me several .blend files."

Willison also describes an operational shift in security practice after auditing Datasette with multiple frontier models: "We'll be incorporating security audits by frontier models into all of our development work going forward." In a separate note quoting Calif Research on a WeChat exploit demo, the security stakes are put starkly: "AI can already do most of the work here."

Taken together, these accounts argue that AI is already changing the work of software production and security, even if the mass-market benefits feel distant. Our role-family figures line up with that picture: infrastructure accounts for 410 open roles, research for 233, and product and design for 95. Those hires support the kinds of pipelines, audits, and tools that Raschka and Willison describe.

Panic, boycott, and the extinction frame

Marcus takes a middle line on existential risk while pressing for accountability. In his boycott case he writes, "Mind you, I happen not to worry too much about extinction. And I don’t think you should be either." In another post about a viral interview, he adds, "I am here to tell you that Anderson Cooper doesn’t need to worry about that particular scenario, and that you don’t either." Alex Hanna, tackling productivity narratives rather than x-risk, is moving the debate to evidence: "My first long-form video is up: Myths Around AI and Productivity!... we dispel 5 common myths around AI and labor by looking at what studies actually say about AI’s impact on labor."

Our hiring data do not arbitrate extinction risk. They do, however, falsify any claim that a boycott or policy panic has frozen employer demand. OpenAI lists 162 open roles, Anthropic 127, Databricks 86, and Waymo 97. Across all employers we see 4,685 open roles, with active 30-day opening rates in the core build categories. On the evidence available to us, the market is staffing up, not stepping back.

Policy heat without a hiring chill

Marcus points to oversight moves he welcomes: "Senator Blumenthal just sent a great letter to OpenAI" and calls out hazards of opacity. Zvi flags legislative extremes. Yet our tracker shows ongoing hiring by the very firms in the crosshairs. Capital One has 190 open roles, Amgen 159, and PwC 128. Whether these moves change the trajectory will take time to show up in postings, but today’s market is not slowing.

What this week establishes

  • Commentary shifted toward fear, boycott, bans, and measurement problems. That is real in the posts, from Mowshowitz’s "preference cascade" to Marcus’s call for a boycott.
  • Employers did not flinch in their hiring. We count 4,685 open AI roles across 178 employers, including 1,912 in modeling and engineering and 846 in data. Accenture, OpenAI, Anthropic, Capital One, and others are actively opening roles in the last 30 days.
  • Safety hiring remains small at 40 open roles, lending weight to concerns about monitorability and risk mitigation not yet being resourced at the same scale as model and product work.
  • Practitioner reports show concrete productivity in creative and security workflows, consistent with Mollick’s diffusion argument and Lambert’s point that average users have not yet felt large, clear benefits.

As always, our headline jobs-created indicator is calibrated to the World Economic Forum’s Future of Jobs Report 2025 (11M created, 9M displaced, by 2030). This week’s postings are a reminder: whatever the temperature of the discourse, employers are still hiring, and mostly for building, not babysitting.

What we read

Every quote above is taken verbatim from one of these posts.

More on this

Today's briefing