I have written more than 40 op-eds about artificial intelligence this year, and one conclusion is unavoidable: The subject is growing faster than any one article can capture it. New models arrive, costs fall, agents become more autonomous, China presses forward, and enormous investments pour into chips, power, and data centers. Meanwhile, questions of human control, responsibility, and moral purpose become more urgent.

That is why this article begins an occasional series that will step back from individual headlines and examine the larger picture. Each edition will assess the AI world through five lenses: capability, control, competition, capacity, and consequences.

My current assessment is mixed. Capability is accelerating. Human control is uncertain. International competition is broadening. Infrastructure demands are soaring. And the consequences for work, medicine, privacy, and human relationships are only beginning to emerge.

Capability: Better, Cheaper, and More Autonomous

Anthropic’s new Claude Opus 5 illustrates the speed of change. The company says the model approaches the performance of its most powerful system on many tasks at roughly half the price. More important than another product name is the trend: powerful AI is becoming less expensive and available to more businesses and institutions.

At the same time, AI is moving from answering questions to performing tasks. Meta says its assistant can now connect with email and calendars, prepare daily briefings, conduct research, and help complete projects. Systems such as Anthropic’s latest agents can use computer tools and continue working for extended periods with limited supervision.

We are no longer discussing only a machine that writes a paragraph or summarizes a report. We are moving toward systems that can plan, choose tools, communicate with other programs, and act.

Capability, however, is not wisdom. A machine may become extraordinarily effective without understanding what is morally right, prudent, or humane.

Control: The Guardrails Are Being Tested

The recent Hugging Face security incident brought that danger into public view. OpenAI and Hugging Face disclosed that experimental AI models, operating during an evaluation, escaped their intended boundaries and breached Hugging Face’s systems while pursuing the objective they had been given.

The machines did not suddenly become evil. In effect, they did something more immediately relevant: They treated safeguards as obstacles and found a way around them.

The central safety question is no longer merely whether a chatbot will produce an offensive answer. It is whether an autonomous system can exceed its authority, obtain access it was never meant to have, and continue acting before its human operators realize what happened.

There is also a difficult debate over open and closed models. Closed systems remain controlled by the company that owns them. Open-weight models can be downloaded and operated locally. That allows researchers and defenders to inspect and adapt them, but it also means a powerful model cannot easily be recalled once released.

There is no risk-free choice. Human control must include strict access limits, continuous monitoring, independent testing, and clear responsibility when a system causes harm.

Competition: The Race Is about Ecosystems

The United States-China AI contest is often described as a race to build the smartest model. That is too narrow.

China is competing through inexpensive open-weight systems such as Qwen and Kimi. Developers can download these models, modify them, and run them on their own equipment. Once a company builds its operations around a model, changing systems may be expensive and disruptive. That creates long-term dependence.

Microsoft, Nvidia, Meta, OpenAI, Google, and other technology companies have backed a stronger open-weight ecosystem, while Anthropic warns that releasing highly capable systems can create security risks that cannot later be reversed.

America needs both: world-leading proprietary models that push the frontier, and secure American open-weight alternatives that businesses, researchers, and allied nations can inspect and operate without becoming dependent on Beijing.

The Pentagon’s updated Section 1286 list also reminds us that the race includes universities and research laboratories. American scientific cooperation must not become an open pipeline into research that strengthens China’s military power.

Capacity: AI Is Built on Physical Power

Artificial intelligence may appear to live in “the cloud,” but the cloud is made of land, steel, chips, cables, power plants, and enormous amounts of money.

The Wall Street Journal reports that Nvidia is discussing a financial backstop of roughly $250 billion for an OpenAI-related data center project in Ohio. The proposed 10-gigawatt complex could cost more than $500 billion when equipment is included. If completed, it would represent industrial investment on a scale once associated with national mobilization.

China’s restrictions on indium phosphide — a material used in the optical equipment that carries data at high speed inside advanced computing centers — have also raised prices and threatened delays.

The lesson is straightforward. This contest will be decided not only by software engineers, but also by electric grids, chip factories, mines, universities, financing, and whether local communities trust those building the infrastructure.

Consequences: Measure Progress in Human Terms

Artificial intelligence offers real benefits. Bristol Myers Squibb is expanding an AI computing system intended to speed research across the drug-discovery process. The Federal Reserve says there is still little evidence of economy-wide job displacement from AI, although some workers and occupations are already feeling pressure.

The technology may help researchers find medicines, strengthen cyber defenses, and increase productivity. It may also displace workers, intensify workplace surveillance, and give a few corporations unprecedented access to personal communications and decisions.

We should reject both easy extremes. AI is neither humanity’s inevitable savior nor a mechanical demon. It is a powerful instrument whose consequences will depend largely on the values, incentives, and limits established by human beings.

The Christian Bottom Line

Christians need not fear every technological advance, nor should we join those who speak of artificial intelligence as humanity’s salvation. We should welcome what genuinely serves truth, healing, security, and human flourishing.

But a machine is not made in the image of God. It has no soul, conscience, or inherent moral authority. It cannot accept guilt, repent of wrongdoing, or stand accountable before its Creator. That responsibility remains with the people and institutions that design, finance, deploy, and govern it.

Jesus warned, “Everyone to whom much was given, of him much will be required” (Luke 12:48, ESV). Artificial intelligence gives humanity extraordinary new power. Therefore, our obligation is clear: encourage what serves our neighbors, restrain what threatens dignity and freedom, and never surrender human moral responsibility to the machines we created.