OpenAI puts a price on safety

OpenAI pays up for risk control

This is a warning disguised as a job post.

OpenAI is hiring a “head of preparedness” with a $555,000 salary plus equity to manage risks tied to mental health, cybersecurity, and misuse. Sam Altman said the role would be stressful and would start in the deep end immediately.

The number matters here. Paying over half a million dollars signals that safety is no longer a side function or a policy memo buried in the org chart. What also stands out is how explicit Altman is about strain, abuse, and models getting harder to control as they improve, which reads less like PR and more like an internal reality leaking out.

This comes after lawsuits, reports of mental health crises, and a 46% jump in companies flagging AI reputational risk in SEC filings. Other AI firms will feel pressure to show similar roles, budgets, and accountability once regulators and courts start asking who was responsible.

If this job is already overwhelming, what does that say about the next models coming online?

AI agents stop being theoretical

This is where AI stops being a demo.

Bernard Marr lays out 10 AI agent platforms that businesses can actually use today, from Google Vertex and Microsoft Copilot Studio to Zapier, Salesforce Agentforce, and QuickBooks’ built-in agents. The point is not new models, but packaging: drag-and-drop builders, deep SaaS integrations, and security layers that let agents touch real workflows instead of toy tasks.

What stood out is how familiar all of this feels. These agents live inside tools people already open every morning, with coffee still warm, which lowers resistance far more than another “AI initiative.” I also think the list quietly shows that differentiation is shifting from intelligence to plumbing, access control, and how safely agents can act across systems.

Once agents can chase invoices, update CRMs, and coordinate thousands of apps through Zapier, experimentation turns into expectation. The risk is teams deploying five agents before agreeing on where one should matter.

If agents are this easy now, what work still needs a human watching the screen?

SoftBank buys the pipes behind AI

This is an infrastructure bet.

SoftBank is acquiring DigitalBridge for $4bn, adding data centers, fiber networks, and cell towers to its growing AI-focused portfolio. DigitalBridge manages about $108bn in assets and will keep operating independently under CEO Marc Ganzi.

What matters here is that Masayoshi Son is buying concrete, steel, and power, not software. When demand for AI spikes, servers hum late at night and electricity meters spin, and Son wants to own the boring assets that everything else depends on. I read this as a calmer, more disciplined move than SoftBank’s past swings.

This fits with Stargate, where SoftBank, OpenAI, and Oracle plan five US computing sites with around 7GW of power. If AI keeps growing, control over capacity becomes leverage.

How many more AI bets end with someone buying land and power first?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.