AI Writes Zero-Day Exploits — Cybersecurity Careers Will Never Be the Same
Google confirmed the first zero-day exploit developed with AI assistance. Attackers used AI to write code targeting an open-source admin tool, aiming for mass exploitation. The exploit was caught before deployment, but Google’s Threat Intelligence Group says this is just the start — hackers are using “persona-driven jailbreaking” to get AI to find vulnerabilities for them.
Why it matters for your career: The cybersecurity job market is about to split in two. On one side: traditional pen-testers and vulnerability researchers whose work AI can now automate or augment. On the other: AI-security specialists who understand both the attack and defence applications of frontier models. If you’re in cybersecurity and you’re not learning how AI models find and exploit vulnerabilities, you’re falling behind. The “AI security engineer” role didn’t exist two years ago. Today it’s the hottest job in the industry.
DeepSeek’s $7B Raise: The Research Lab Becomes a Company
DeepSeek is seeking up to US$7.35 billion in its first external funding round. The company that proved you could train competitive models cheaply is now going commercial. DeepSeek V4.1 is reportedly due in June, with founder Liang Wenfeng and China’s state semiconductor fund backing the raise.
Why it matters for your career: DeepSeek going commercial means more jobs in the Chinese AI ecosystem — but also more competition for AI roles globally. When a $7B-funded lab starts hiring, it pulls talent from everywhere. The bigger signal: the “research-first” phase of AI labs is ending. Every major lab is now a product company. If you were hoping to join an AI lab for pure research, that window is closing. Product-minded AI engineers are the future.
China’s AI Agent Rules: “Humans Must Stay in the Loop”
China’s draft regulations for AI agents require that humans retain the ability to review and override agent decisions. The rules distinguish between decisions limited to users, decisions requiring user authorisation, and autonomous decisions. Potential agent tasks identified in the draft range from marking homework to managing “the entire bidding and tendering process.”
Why it matters for your career: If you’re building AI agents, the “human in the loop” requirement is going to be a design constraint everywhere — not just China. The EU and US will likely follow. This means roles like “AI agent governance specialist,” “human-AI interaction designer,” and “agent oversight engineer” are about to become real job titles. Companies deploying agents without human oversight mechanisms will face regulatory risk.
Baidu’s 6% Training Cost: What It Means for AI Teams
Baidu claims Ernie 5.1 achieves competitive performance at just 6% of the industry’s pre-training cost. If accurate, this means mid-sized companies with GPU clusters can now train competitive models — not just Big Tech.
Why it matters for your career: Cheaper training means more AI teams, not fewer. When a mid-sized NZ company can fine-tune a competitive model without a $100M compute budget, the demand for AI engineers expands dramatically. But it also means the barrier to entry drops — and your competition increases. Specialisation matters more than ever. “General AI engineer” is becoming a crowded field. “AI engineer who understands [your industry]” is where the money is.
Alibaba’s Agentic Shopping: New Roles in AI Commerce
Alibaba integrated Qwen AI with Taobao’s 4 billion+ items and Alipay-native checkout, creating the largest agentic commerce platform. An AI agent can now search, compare, and purchase products with a single conversational prompt.
Why it matters for your career: E-commerce companies need people who understand agentic commerce — not just chatbot customer service. Product managers who can design agent-native shopping flows, engineers who can build agent-to-marketplace APIs, and trust & safety specialists who can prevent agent-based fraud are all about to be in demand. If you’re in e-commerce and not thinking about AI agents as your primary customer interface, you’re planning for 2024.
🔍 THE BOTTOM LINE
This week’s career signal is clear: the AI industry is shifting from “build the model” to “deploy the agent.” DeepSeek goes commercial. China regulates agents. Alibaba ships agentic shopping. Baidu makes model training cheap enough for everyone. The jobs are moving from the lab to the product — and from the model to the agent. If you’re positioning yourself for the next phase of AI, ask not “how do I build a model?” but “how do I deploy an agent safely, responsibly, and at scale?”