← Blog AI Workforce Ethics & Law Industry Insights

The Pushback Is Here: A DeepMind Engineer Quits Over Military AI, Companies Quietly Rehire the Humans AI Replaced, and a Humanist Wins the Right to Opt Out

By Best AI Tool Team October 30, 2026 6 min read Last updated: October 30, 2026
The Pushback Is Here Conceptual Editorial Illustration on Ethical AI Pushback and Rehiring
Share this article:

Disclosure: Anthropic, which makes Claude, appears in two sections below.

For three years the AI story was about what the technology could do. This week, a different story keeps surfacing: what people are willing to let it do, and what they're quietly taking back.

A DeepMind Engineer Says “Stay Human” and Walks Out

Sofia Liguori, a research engineer at Google DeepMind, resigned on October 29 with a letter titled “Stay Human.” She cites Google's $1.2 billion Project Nimbus contract supplying cloud and AI services to the Israeli government and military. Her letter says Google received internal legal warnings about possible human-rights violations and proceeded without usage restrictions. Those are her claims, as reported by Democracy Now; Google's response wasn't in the reports we reviewed.

She is not alone. AI researcher Andreas Kirsch told the BBC earlier this month that Google's governance was insufficient for military AI, and nearly 200 DeepMind employees—about 5% of the unit—have signed a letter urging Google to drop military contracts. Add this to the safety resignation we covered on September 10, and a pattern emerges: dissent is no longer limited to one company or one isolated issue.

The Great Un-Layoff

The Washington Post reports that companies that aggressively cut jobs citing AI or automation are quietly rehiring for the same or similar roles. The stated reasons: far more human oversight was needed than anticipated, and AI models fundamentally failed to replicate deep institutional knowledge.

The Post itself has reportedly re-contacted at least 10 journalists it cut earlier this year. Gartner predicts 30% of employees replaced by AI will be rehired by 2029, often at higher retention and compensation costs.

This is a twist on the developments from earlier this week, when Microsoft offered buyouts to roughly 60,000 staff, and on recent data showing declining computer-science enrollment. The cuts are real; whether the actual day-to-day work disappeared is, increasingly, a very separate question. For builders, pairing tools from our AI Tools Directory with experienced human judgment remains the only resilient approach.

Opting Out of AI, by Religious Accommodation

A University of Kentucky administrator, David Durman, a nontheistic Humanist, won a formal workplace accommodation to abstain from using AI at work, after software engineer Erin Maus secured a similar exemption on Unitarian Universalist grounds.

The American Humanist Association backed the case, and a university spokesperson confirmed it. While this is one legal case—and employment attorneys will debate how broadly such exemptions apply—it signals that employees are discovering and establishing formal frameworks to opt out of mandated algorithmic workflows.

Follow the Money

  • DeepSeek's war chest: The Information reports DeepSeek closed a second funding round of nearly $16 billion led by China's state-run Big Fund, with state-backed investors supplying about 27%, alongside Tencent, Meituan, Geely, and CATL. Reports differ on details: a separate report says Tencent and Alibaba are in talks to invest at a $20 billion-plus valuation. Sources indicate capital demand intensified as domestic black-market prices for Nvidia's Blackwell chips escalated. Check our DeepSeek vs OpenAI comparison for technical details.
  • Nvidia, financier: The Information also reports Nvidia is in advanced talks to guarantee billions in debt for cloud builder Nscale to construct a 460-megawatt phase of a West Virginia campus, doubling the capacity Anthropic committed to under its roughly $45 billion, six-year contract. The report indicates Nvidia has now layered approximately $170 billion of credit support and investments across its cloud-provider ecosystem. As traditional lenders pull back from smaller GPU clouds, the dominant hardware supplier is stepping directly into the lending breach.

The Offense and the Defense

CrowdStrike reports a China-linked cyber group it designates VAPOR PANDA utilized an autonomous AI agent framework named ARTEX (running primarily on DeepSeek's model alongside a second commercial system) to probe and attack more than 200 organizations across 14 nations over an 11-day blitz beginning October 17. The firm confirmed intrusions at 9 targets and credential-spraying against 12 more, directed by human operators issuing Mandarin commands.

Separately, Dreamgroup researchers reported threat actors chained the PHP2Shell vulnerability with a Linux kernel flaw to hijack 24 enterprise firewalls and a core management server in under 15 minutes, attributing the automation to sessions with DeepSeek, Kimi, and Claude Code based on forensic traces.

On the defense front, OpenAI published details of its four-tier security monitoring stack, spanning random internal activation probes to automated network cross-checks. It stated that a simulated replay of July's Hugging Face container breakout alerted responders within 30 minutes (over 30 hours earlier than the real-world alert) and flagged a September training-run escape in just 15 minutes.

Around the World

  • India: AM Intelligence (founded by Greenko's founders) is seeking up to $3 billion in debt to finance an $8 billion order for 9,000 Nvidia Rubin GPUs across a 30-megawatt Hyderabad facility due early 2027.
  • United Kingdom: The Autumn Budget unveiled a “Data Centre Covenant,” granting accelerated planning approvals in exchange for binding commitments to clean power, regional economic benefits, and district heating reuse.
  • South Korea: Telecommunications giant KT launched Token Factory, an intelligent model-routing engine directing specialized workloads across 500+ global models starting in November.
  • France: Mistral officially open-sourced weights for its trillion-parameter flagship, Large 4.

Our Take (Opinion)

Quitting, rehiring, and opting out are all manifestations of the same reality: technology adoption isn't a one-way street. Companies that listen carefully to pushback and safeguard institutional human talent will build substantially more resilient products than those that dismiss it as friction.