AI Hiring Brief · 2026-07-14

19 frontier AI labs · 66 roles added in 24h · agihunt.info/en/jobs

Today at a glance

OpenAI doubles down on Codex enterprise security: Four of today's eight new roles point squarely at the Codex stack — two engineering roles in Applied AI and two product managers in Security Products, covering security controls, partner interfaces, and a dedicated Cyber Defense and Blue Team function. The compliance scaffolding around an AI coding agent is now as prominent as the agent itself.
Microsoft AI leads the week by a wide margin: +26 today, +148 over seven days — the largest single-week addition across all 19 companies tracked. Roles span firmware validation, Agent 365 product management, forward-deployed engineering in the UK, Netherlands, and Switzerland, and AI workforce specialists in Japan and Brazil. This is coordinated global enterprise rollout, not a hiring spike.
Google DeepMind stretches into hardware prototyping and biosafety: A "Lead Prototyping Machinist" in Cambridge, MA and a "Research Scientist, Robotics" in London confirm that DeepMind's robotics work has moved past pure research into physical fabrication. A "Principal Virologist, Responsible Development and Innovation" posted across six US cities is the day's most unusual signal — biological risk assessment is becoming a standard pillar inside frontier lab safety teams.
Anthropic plants flags in Seoul and Paris: An "Enterprise Account Executive - FSI" in South Korea and "Startup Partnerships - France & Southern Europe" in Paris landed on the same day, extending Anthropic's direct sales reach into two geographies where it had no visible presence before. The FSI (financial services) label for Seoul marks one of the most compliance-sensitive enterprise verticals.
ModelBest pivots toward agent infrastructure: The Beijing-based lab added four roles in a single day, including an "Agentic Engineering" position in its 基础架构 (infrastructure) department — a small company making a concentrated bet on agent runtime engineering.

Company moves

OpenAI (+8 today, +84 over 7 days) opened four Codex-facing roles: "Software Engineer, Codex — User Activation" and "Software Engineer, Codex — Enterprise Controls" in Applied AI, plus "Product Manager, Codex Security Controls & Partner Interfaces" and "Product Manager, Cyber Defense and Blue Team" in Security Products. The security products department now has two PMs focused on Codex alone. Three User Operations roles — "Support Operations Manager," "Product Engagement Specialist, User Operations," and "Product Engagement Manager, User Operations" — signal a parallel push to build the operational layer beneath the product.

Anthropic (+3 today, +61 over 7 days) added a "Staff+ Software Engineer, Safeguards Review Tooling" in AI Research & Engineering, keeping its safety tooling investment visible. The two sales roles — "Enterprise Account Executive - FSI" in Seoul and "Startup Partnerships - France & Southern Europe" in Paris — represent concrete geographic expansion into markets Anthropic has not publicly staffed before.

Google DeepMind (+4 today, −13 closed, net negative for the day) added a "Senior Software Engineer, Android, GeminiApp Experience" for the consumer Gemini product, a "Lead Prototyping Machinist" tied to physical robotics development, a "Research Scientist, Robotics" in London, and the aforementioned virologist. The 13 closures against 4 additions means the overall open count contracted on the day — worth noting given the week's +37 figure.

Microsoft AI (+26 today, +148 over 7 days, 38 closed) generated the broadest job list of any company. Highlights include "Principal Product Manager (Agent 365)" in Dublin, "Principal Firmware Validation Engineer" spanning Redmond, Mountain View, Austin, Hillsboro, and Raleigh, "AI Business Solutions Workforce Specialist (M365, Copilot & AI Agent solution, 大手エンタープライズ企業担当）" in Tokyo, "AI Workforce Digital Specialist Manager" in São Paulo, and "Senior Research Software Development Engineer, MSR AI for Science" in Germany and the Netherlands. A "Researcher" role opened in Beijing. The volume and distribution point to an organization in active global deployment of its AI products, not a planning phase.

Meta AI (+10 today, +22 over 7 days) posted a "Content Engineer, Meta Superintelligence Labs" in Menlo Park — among the first externally visible role attached to that newly named unit. Also notable: "Software Engineer, Systems ML - Compilers / Backend" in Sunnyvale (inference stack depth), "Research Scientist, Omni-lingual (FAIR)" in Paris (multilingual research continuing at FAIR), and "Supply Chain Operations Lead, Infrastructure Supply Chain" in Austin (data center logistics build-out).

NVIDIA (+5 today, +44 over 7 days) added a "Senior System Software Engineer - AI Data Platform - Inference Factory Optimization" — the phrase "inference factory" in a job title is notable; it signals that optimizing inference at scale is now a named discipline internally. A "Senior Accelerated Computing Architect" in Shanghai and a "Backend Compiler Engineer" across six locations continue NVIDIA's sustained compiler and inference investment.

ModelBest (面壁智能, +4 today, +10 over 7 days) posted "算子工程师" (kernel/operator engineer), "Agent算法工程师" (agent algorithm engineer), "科研项目经理" (research project manager), and "Agentic Engineering" in infrastructure. For a lab of this size, four roles in one day concentrated on agent engineering and underlying operator development is a clear signal of a strategic build-out.

Baidu AI (百度 AI, +2 today, +14 over 7 days, 23 closed) added "具身智能数据服务解决方案与项目经理（J102453）" (embodied intelligence data services solution and project manager) in its ACG department — one of the more explicit embodied AI commercialization roles visible among Chinese labs this week. The 23 closures against 2 additions represents the day's sharpest single-company contraction.

Lovable (+3 today) opened three growth-oriented roles: "Lifecycle Marketer, Self-Serve," "Performance Marketer," and "Product Experience Specialist (Full Time)" across London and Stockholm. The self-serve framing suggests the company is leaning on a paid conversion funnel rather than direct sales as its primary growth mechanism.

Cursor (+1 today) added "Technical Program Manager, Field" in Customer Success — a sign that enterprise deployments have grown complex enough to warrant dedicated field delivery management.

Apple AIML, SpaceXAI, Perplexity, Moonshot AI (月之暗面 Kimi), DeepSeek (深度求索), Zhipu AI (智谱 AI), StepFun (阶跃星辰), and Meituan LongCat (美团·龙猫) posted no new roles today. Mistral AI recorded zero additions for both the day and the full seven-day window.

What it signals

The seven-day figures reveal a significant gap between the most active global labs and the rest. Microsoft AI (+148), OpenAI (+84), and Anthropic (+61) account for the overwhelming majority of additions among global companies, while most of the Chinese labs tracked — DeepSeek, Zhipu AI at zero for the week; StepFun at +6, Moonshot AI at +1 — are in a hiring pause. This divergence could reflect different fiscal calendars, different stages of funding cycles, or a deliberate consolidation after earlier waves of expansion. It is consistent with what has been visible across multiple recent reporting windows.

The concentration of OpenAI's new roles inside a security products department built around Codex is a structural signal, not just a daily fluctuation. Enterprise buyers in regulated industries — financial services, healthcare, government — require audit trails, permission scopes, and red-team documentation before deploying AI in production workflows. The simultaneous appearance of "Codex Security Controls & Partner Interfaces" and "Cyber Defense and Blue Team" on the same day, alongside Cursor's field TPM hire, suggests that the competitive battleground for AI coding tools is shifting from benchmark performance toward compliance infrastructure. Capability gaps between leading coding agents are narrowing; compliance gaps are where contracts are won or lost.

Google DeepMind's "Principal Virologist, Responsible Development and Innovation" continues a trend that has been quietly building across frontier labs: the internalization of domain-specific scientific expertise into safety functions. A virologist sitting inside a responsible AI department is not evaluating model outputs — they are assessing what happens when a capable language model intersects with dangerous biological knowledge. This is a form of pre-competitive safety investment, the cost of which no single lab wants to absorb but all are now under pressure to demonstrate. The same session that added a virologist also added a prototyping machinist in Cambridge, MA — physical robot fabrication and biosafety risk assessment, coexisting in one lab's weekly additions, illustrate how wide DeepMind's research perimeter has become.

Microsoft AI's +148 over seven days is the figure that most clearly captures where AI investment is concentrating right now: not in research headcount, but in the global machinery needed to put AI products in front of enterprise customers. Agent 365 product management in Dublin, AI workforce specialists in Tokyo and São Paulo, forward-deployed engineers across Western Europe — this is a distribution build, not a capability build. When a company adds firmware validation engineers at five US sites in the same week it opens enterprise sales roles on three continents, it is signaling that infrastructure and market access are the two constraints it is trying to relieve simultaneously, and that it believes the model capability question is largely settled for its near-term product roadmap.