In the Army, I learned that every operation begins with a hard look at the enemy. Before a commander approves a mission, the staff must explain the opposing force’s capabilities, intentions, and likely courses of action. Get the threat wrong, and everything that follows rests on a false assumption.

Corporations run their own version of that drill. They study competitors, anticipate emerging markets, and recruit the people tomorrow will require. A company’s payroll can reveal its real expectations more clearly than its public relations department.

That makes Anthropic’s current hiring spree more than a personnel matter. The company behind the Claude family of AI models is assembling teams to confront weapons proliferation, cyber conflict, foreign influence, surveillance, and economic upheaval — all fields it plainly expects to matter as AI capabilities advance.

History shows why hiring patterns matter. Bell Labs assembled about 4,000 scientists in 1925 to pursue research expected to shape communications; two decades later, its researchers produced the transistor. Volvo recruited Nils Bohlin as its first safety engineer in 1958. Within a year, he developed the three-point seat belt, making safety central to Volvo’s identity. In 2002, Microsoft halted new work by 8,500 Windows Server developers and spent more than $200 million retraining them in security, hiring for the future it saw coming and building trust around the risks that future would bring.

Anthropic appears to be following all three patterns at once: developing systems that could create vast new markets, recruiting specialists to contain the dangers those systems may produce and building the trust needed to sell them to government and industry. Its openings amount to a corporate threat assessment, touching weapons of mass destruction, cyberwarfare, foreign influence, surveillance, model theft, and economic disruption.

Start with weapons. Consider the advertised Threat Intelligence Manager for CBRN-E and Advanced Weapons. CBRN-E is government shorthand for chemical, biological, radiological, nuclear, and explosive threats. Anthropic wants this manager to build a team detecting attempts to use its systems to develop, improve, or distribute such weapons, with backgrounds in counterproliferation and weapons-of-mass-destruction analysis preferred. An active Top Secret clearance is desirable, and the salary reaches $455,000 a year. Companies do not offer nearly half-million-dollar salaries to prepare for dangers they consider imaginary.

The worry is not that a chatbot decides on its own to build a bioweapon. The nearer danger is that capable systems could shrink the expertise, manpower, and time a terrorist group or technically skilled individual needs to attempt something previously out of reach. The same system that accelerates drug development may help a malicious user navigate a dangerous chemical or biological problem. Anthropic profits by making its systems more capable, yet cannot sell those capabilities unless customers believe they can be controlled. Safety becomes a requirement for growth, not just a moral obligation.

The same pattern shows up in cybersecurity. Anthropic is standing up a new team under a Frontier Red Team cyber lead. The posting says frontier models will have “dramatic impacts” on cybersecurity and that significant portions of the field may have to be fundamentally reconsidered, including the use of autonomous systems to find and repair vulnerabilities and simulations pitting machine defenders against machine attackers.

We are moving from tools that merely advise a hacker toward systems that search for vulnerabilities, write code, test attacks, and run multistep operations with less human supervision. The same tools sold to developers rely on capabilities also sought by spies, criminals, and foreign militaries. The industry may end up selling both the engine of automation and the armor required to survive it.

Political warfare offers another warning. Anthropic wants a Threat Intelligence Manager for Influence Operations and Surveillance to build a team investigating state-linked manipulation, coordinated false identities, commercial spyware, and surveillance conducted for authoritarian governments. The company specifically seeks experience with Chinese, Russian, and Iranian state-sponsored operations.

Propaganda once addressed the crowd. Artificial intelligence can whisper separately into every ear. Such a system could generate thousands of convincing identities, speak local languages, study individual users, and tailor arguments to their fears and desires, cultivating a relationship for months without tiring or breaking character. That is more than spreading false information. It is manipulation at a personal scale, striking at the freedom of conscience on which responsible citizenship depends.

Anthropic is also seeking a Transformative AI Research Economist to model economic conditions that may have no historical precedent, examining growth, employment, income distribution, and economic governance under different scenarios. That posting tells us Anthropic is not preparing merely to make office workers a little more efficient. It is preparing for a technology capable of restructuring labor markets and concentrating enormous wealth and influence.

This should matter especially to Christians. Work gives structure to daily life and a meaningful avenue for service and stewardship. Human dignity, however, does not depend upon economic productivity; it comes from being made in God’s image, as Genesis 1:27 declares. An economy that treats people as inefficient production costs misunderstands both work and the worker.

Anthropic’s strategy carries a troubling paradox. The company may sincerely want to prevent catastrophe; its Responsible Scaling Policy acknowledges that dangerous capabilities not present today could emerge quickly as AI systems advance. Yet Anthropic also needs governments, customers, and the public to trust it enough to keep developing the systems producing those concerns.

In my new book, “The Final Algorithm,” I warn that societies court disaster by trusting systems they built more than the One who built them. A corporation grading its own safety exam is no exception.

Safety can become a competitive advantage, securing government partnerships and setting Claude apart from rivals, while expensive compliance favors large corporations and raises the barrier for smaller ones. The same corporation builds the capability, defines the danger, tests the product, judges its own safeguards, and profits when deployment proceeds. The public bears the consequences if that judgment proves wrong. Good intentions are not a system of checks and balances.

Congress should treat these openings as an early warning, not an employment curiosity.

America needs independent testing of frontier systems, stronger protection against model theft, clear liability for foreseeable damage, and firm limits on delegating lethal force, surveillance, or critical infrastructure to AI systems. Above all, human responsibility must remain clear. No executive, programmer, or official should get to blame “the algorithm” for decisions that harm people.

Scripture assigns real accountability to the one who spots danger and stays silent. Ezekiel 33:6 warns that when a watchman sees the sword coming and fails to sound the alarm, God says “his blood I will require at the watchman’s hand.” Anthropic has sounded its trumpet. The rest of us have no excuse for not listening.

Anthropic sees danger ahead. Its help-wanted page is more than a list of vacancies. It is a warning order from one of the corporations building the future. The rest of us would be foolish not to conduct a threat assessment of our own.