The last 7 days were basically the AI industry asking: “What if every part of the stack became a hiring problem at once?”

OpenAI flagged possible “critical” cybersecurity risk in its upcoming Astra model and paused parts of development. Alibaba is reportedly planning to charge large commercial users of its next open-weight Qwen model through revenue sharing. Big Tech’s AI data-centre race has created more than $1 trillion in future lease commitments. Anthropic is building an in-house chip design team. And Etsy cut 12% of its workforce, mostly in product and engineering, because apparently even handmade marketplaces are not safe from the great operational haircut.

The hiring signal is clear: AI is no longer just creating model roles. It is creating new demand in cyber safety, AI governance, inference economics, infrastructure finance, chip design, and people who can explain the difference between “AI-powered” and “actually useful.” The bar is low. The job specs are somehow still worse.

The Drop

1) OpenAI pauses work on Astra after possible “critical” cyber risk

What happened: OpenAI flagged possible critical cybersecurity risk in its upcoming model, Astra, saying it may have the ability to autonomously find and exploit zero-day software vulnerabilities. OpenAI paused parts of internal development and triggered stronger safety protocols.

Why it matters for hiring: This is a major escalation. Model safety is no longer only about hallucinations, bias or policy compliance. It is now about whether a model can independently discover and exploit real vulnerabilities.

Roles likely to spike:

• AI security engineers

• Cyber eval engineers

• Model safety researchers

• AppSec and vulnerability specialists

• AI red-team engineers

• Secure deployment and sandboxing engineers

Hiring takeaway: Any company building or deploying frontier models now needs a serious cyber-safety function. Not a token “security review.” A real function with evals, containment, monitoring, escalation and incident response.

2) Alibaba plans to charge major users of its next open-weight Qwen model

What happened: Alibaba reportedly plans to ask large commercial users of its next Qwen open-weight model for a share of revenue generated from the model. The model, Qwen3.8-Max, is expected to follow the wave of high-performing Chinese open-weight models competing with US frontier labs.

Why it matters for hiring: The “open model” market is maturing fast. Open-weight does not automatically mean free, simple or risk-free. Companies building on open models need people who understand licensing, hosting, monetisation, evaluation, cost and compliance.

Roles likely to spike:

• Open-model deployment engineers

• Model evaluation specialists

• AI platform engineers

• Legal / commercial AI product leads

• Inference optimisation engineers

• Multi-model architecture specialists

Hiring takeaway: Open models are becoming commercial infrastructure. Teams need to stop treating them as cheap backup and start treating them as part of enterprise architecture.

3) Big Tech’s AI buildout creates a $1.09T future lease burden

What happened: Microsoft, Meta, Oracle, Amazon and Alphabet have committed about $1.09T in future payments under leases that have not yet begun, mostly for data centres needed to support AI. These commitments are not yet showing up as full debt-like lease liabilities on company balance sheets.

Why it matters for hiring: AI infrastructure spending is not just capex. It is now a long-term financial commitment that will shape hiring, cost control and operating discipline for years.

Roles likely to spike:

• AI infrastructure finance

• Data-centre procurement

• FinOps and capacity planning

• Power and grid strategy

• GPU cluster operations

• Infrastructure programme management

Hiring takeaway: Senior AI infrastructure roles should now include financial accountability. If a team cannot explain utilisation, cost-per-workflow, reserved capacity, energy exposure and vendor lock-in, it is not ready to scale AI. It is just expensive with confidence.

4) Anthropic builds an in-house chip design team for Claude

What happened: Anthropic confirmed it is building an in-house chip design team for Claude, hiring engineers across hardware and software to co-design chips and AI models. The company will continue using hardware from AWS, Google, Nvidia and AMD, but wants more control over performance and scalability.

Why it matters for hiring: Model companies are moving deeper into hardware. The best AI companies do not want to be stuck waiting for chip supply, power capacity, cloud availability or someone else’s roadmap.

Roles likely to spike:

• AI chip design engineers

• Hardware / software co-design specialists

• Compiler and runtime engineers

• ML systems engineers

• Performance and efficiency engineers

• Silicon programme managers

Hiring takeaway: AI hiring is drifting toward the full stack: model, runtime, chip, memory, networking, power and cost. The recruiter who can map that stack will beat the recruiter searching “LLM engineer” and hoping.

5) Etsy cuts 12% of workforce, mostly product and engineering

What happened: Etsy laid off around 12% of its workforce, about 220 employees, mostly in product and engineering, as part of a restructuring plan. The company said affected employees will receive at least 16 weeks of severance and healthcare support.

Why it matters for hiring: This is the clearest workforce signal this week outside pure AI. Product and engineering teams are still being reshaped, even in companies that are not frontier AI labs. The market is still selective, cost-conscious and under pressure to prove productivity.

Roles likely to be available:

• Product engineers

• Marketplace engineers

• Mobile engineers

• Data analysts

• Product managers

• Engineering managers

Roles likely to stay competitive:

• AI search and recommendation engineers

• Growth engineering

• Marketplace trust and safety

• Payments / risk engineering

• Infra and reliability

Hiring takeaway: Good product and engineering talent is still entering the market, but companies will prioritise people who can connect product work to revenue, automation, trust, marketplace quality or operational efficiency.

Smaller Company Watch

Obsidian Security | $85M Series D | $1.1B valuation

Obsidian raised $85M as demand rises for securing AI agents across enterprise apps. Nearly 70% of its clients reportedly now allow AI agents to interact with business data.

Likely hires:

• AI agent security engineers

• SaaS security engineers

• Cloud security specialists

• Product security

• Enterprise security GTM

Why it matters: The AI agent security market is becoming a proper category. When agents can access Salesforce, Microsoft Copilot Studio, Claude and internal data, security becomes more than “please don’t paste secrets into chat.”

Volta Infra | $2.4B valuation | $10B AI cloud partnership

Volta, founded just seven months ago, announced a $10B AI cloud computing partnership in Europe and a $5B AI infrastructure initiative with asset manager Azora.

Likely hires:

• Cloud infrastructure engineers

• European data-centre operations

• Power and cooling specialists

• Capacity planning

• Enterprise cloud sales

Why it matters: New AI infrastructure companies are scaling at ridiculous speed. Good for hiring demand. Bad for anyone pretending “cloud” is still just AWS credits and optimism.

AMD acquires Taalas to deepen AI inference chip push

AMD acquired Taalas, a startup founded in 2023 that had raised around $219M in total funding for chips optimised for AI models.

Likely hires:

• AI inference chip engineers

• Hardware / software co-design

• Compiler engineers

• Silicon architecture

• Performance engineering

Why it matters: Nvidia challengers are still buying teams, not just roadmaps. Inference efficiency is now a talent market.

DeepSeek invests $20.8M in Unitree’s Shanghai IPO

DeepSeek invested $20.8M in robotics company Unitree, taking a 2.31% stake and signalling deeper interest in embodied intelligence.

Likely hires:

• Robotics software engineers

• Motion-control engineers

• Multimodal AI specialists

• Simulation engineers

• Physical-world data engineers

Why it matters: AI is moving from chat boxes into robots. Extremely normal. We taught language models to reason, and now someone is giving them legs.

Siemens Energy posts record quarter on AI data-centre power demand

Siemens Energy reported record third-quarter sales, margins and orders, driven partly by AI data-centre demand in the US and power projects in the Middle East.

Likely hires:

• Grid and power systems engineers

• Gas turbine specialists

• Data-centre energy strategy

• Project engineers

• Infrastructure commercial leads

Why it matters: AI hiring is no longer just software. Power engineering is becoming part of the AI stack.

AI Tool of the Week

Loxo

What it does: Loxo is an AI recruiting platform that combines sourcing, CRM, ATS, outreach, people data and AI agents into one system. It also describes an 800-million-person talent graph and tools for candidate rediscovery and multi-channel outreach.

Who it’s for: Recruitment agencies or lean internal hiring teams that want sourcing, CRM and outreach in one place rather than a Frankenstein stack of LinkedIn, spreadsheets, email sequencers and silent despair.

Quick pilot idea this week:

• Pick one hard-to-fill AI security, platform or infrastructure role.

• Define 6 must-have criteria and 4 “strong signal” criteria.

• Use Loxo to build a 50-person shortlist from both new sourcing and CRM rediscovery.

• Manually audit the top 20 and bottom 20.

• Compare against a fresh LinkedIn Recruiter-only search.

Metrics to track:

• Relevant profiles per hour

• % of shortlist from existing CRM

• Outreach response rate

• Hiring manager approval rate

• Screen-to-interview conversion

• Time-to-first-qualified-shortlist

Why this tool fits this week: The market is getting more specialised. If you are hiring AI security, chip, infra or open-model talent, you need better rediscovery and signal-matching. Keyword-only sourcing is basically using a spoon to tunnel out of prison.

Hiring / Interview Insight

Test for “AI risk ownership,” not just AI enthusiasm

Astra’s cyber-risk pause, the EU AI Act, rogue-agent incidents and Obsidian’s agent-security raise all point in one direction: companies need people who can own AI risk inside real workflows.

Add a 30-minute AI risk ownership station:

Give candidates this scenario:

“You are deploying an AI agent into a business workflow. It can access internal tools, summarise sensitive documents, generate customer-facing content, and trigger actions in production systems. A new model version shows stronger cyber capability than expected. What do you pause, test, restrict, log, communicate or roll back?”

Score for:

• Risk triage

• Access control thinking

• Cyber-safety awareness

• Model evaluation design

• Logging and auditability

• Customer / stakeholder communication

• Human escalation and rollback planning

Metrics to track:

• Pass-through rate by seniority

• Quality of risk reasoning

• New-hire 60-day manager satisfaction

• AI incident rate

• Time-to-containment for AI workflow errors

• Number of AI workflows with named owners

Why it matters: The strongest candidates will not be the ones who say “AI can do everything.” They will be the ones who know exactly where it should not be allowed to touch anything yet.

Funding Watch

Obsidian Security | $85M Series D | $1.1B valuation

Hiring signal: AI agent security, SaaS security, enterprise governance, cloud controls.

Volta Infra | $2.4B valuation | $10B AI cloud partnership

Hiring signal: European AI cloud, data-centre operations, power, compute infrastructure.

AMD / Taalas | acquisition of AI inference chip startup

Hiring signal: inference chips, silicon architecture, compiler/runtime, hardware/software co-design.

DeepSeek / Unitree | $20.8M strategic IPO investment

Hiring signal: embodied AI, robotics, motion control, simulation, physical-world data.

Nvidia / Lancium | reported up to $3B investment

Hiring signal: power infrastructure, Stargate-adjacent data centres, grid access, AI compute buildout.

Siemens Energy | record quarter

Hiring signal: AI power supply chain, gas turbines, power project delivery, data-centre energy.

Quick Bytes

• The Trump administration is reportedly drafting a ban on Chinese data-centre devices, adding another supply-chain risk layer to AI infrastructure planning.

• Oracle’s AI strategy is creating a high-stakes ratings gamble as the company leans further into infrastructure commitments.

• Google shook up AI leadership, with Demis Hassabis shifting into a chief scientist / chairman role and Koray Kavukcuoglu taking stronger operational leadership.

• Anthropic named Mariano-Florentino Cuéllar as its first chief global affairs officer, another sign that AI policy is now part of product strategy.

• Apple is reportedly testing China’s CXMT memory chips for iPhones and MacBooks as the AI memory squeeze continues.

• Corporate AI adoption still looks messier than the sales decks suggest, with Breakingviews noting that corporate buyers are under pressure to prove actual financial returns.

What to do this week

1) Add AI risk ownership to senior interviews

Target roles: AI product, platform, security, backend, solutions.
Metric: every senior candidate handles one scenario involving model risk, restriction or rollback.
Why: Astra’s pause shows model capability can move faster than deployment confidence.

2) Build a Qwen / open-model commercial-risk checklist

Target areas: licensing, revenue share, hosting, data controls, evals, fallback models.
Metric: every open-model deployment has commercial, legal and technical owner sign-off.
Why: Alibaba’s reported Qwen monetisation plan shows open-weight does not mean consequence-free.

3) Add lease and capacity awareness to AI infra hiring

Target roles: AI infra, FinOps, platform leads, data-centre strategy.
Metric: every AI infra spec includes cost, capacity, utilisation or vendor commitment ownership.
Why: Big Tech’s $1.09T lease commitments show infrastructure cost is a hiring problem now.

4) Map embodied AI and chip-inference talent

Target profiles: robotics, chiplets, inference chips, motion control, simulation, compiler/runtime.
Metric: 30 qualified profiles added this week.
Why: DeepSeek/Unitree and AMD/Taalas show the AI stack is moving into physical systems and specialised inference hardware.

Outro

This week’s takeaway is simple: AI hiring is becoming less about who can use the newest model and more about who can control the risk, cost, infrastructure and commercial model around it. The winners will hire people who understand cyber safety, open-model economics, data-centre commitments, chip constraints and workflow ownership. The losers will keep posting “AI-native engineer” and wonder why the shortlist looks like a prompt-engineered tragedy.

That’s all for this week’s Tech Talent Drop — stay informed, and see you next week!