Month: September 2026

How Scary Is AI Getting?
The machines are not plotting our extinction in a secret bunker. The truth may be more complicated – and more troubling – than that.

WIRED examines why recursive self-improvement, agentic swarms and recent security incidents are alarming researchers inside leading AI labs.

There is an old adage in writing: you cannot make this stuff up.

Of course, fiction writers make things up all the time. Cryptogeddon comes from my imagination. Its characters, conspiracies and crises are inventions. But the world in which I am writing it keeps producing developments that would have sounded implausible when I began.

Over the past few weeks, the tone of the AI conversation has shifted again. Not because another chatbot can write a better email or produce a prettier picture, but because people working near the frontier are sounding genuinely frightened by what they are building.

A recent WIRED report describes researchers worried about recursive self-improvement, autonomous agents and a loss of meaningful human control. One senior safety researcher put the possibility of AI killing everyone at greater than ten per cent within the next decade.

That is one person’s estimate, not a scientific measurement or a consensus forecast. But it is not coming from a stranger on the street wearing an “END IS NIGH” sandwich board. It is coming from inside the industry.

So: how scared should we be?

The honest answer is neither not at all nor we are all doomed.

We should be scared enough to pay attention.

Why This Looks Different From Where I Sit

My work with Cisco AI Defense puts me inside the challenge of securing the models, applications, data and infrastructure that make up an AI system.

I have spent much of my professional life thinking about what happens when technology, people and risk collide.

I have worked in cybersecurity for decades, including serving as a CISO. Today, my work with Cisco AI Defense puts me deep inside the security problems created when organizations connect increasingly capable AI systems to their data, applications and infrastructure.

That experience changes how I read stories about AI risk.

My first question is not whether a machine has become conscious, malevolent or secretly ambitious. It is usually much less cinematic:
- What access was it given?
- What objective was it pursuing?
- How was that objective bounded?
- What assumptions did the people deploying it make?
- What happened when one control failed?
- Was anyone actually watching what the system did next?
Cybersecurity has taught us repeatedly that catastrophic outcomes rarely require a single, all-powerful adversary. They emerge from combinations: excessive permissions, misunderstood dependencies, poor configuration, weak monitoring, competitive pressure and humans assuming that someone else has accounted for the risk.

AI does not replace those familiar problems.

It adds speed, scale, autonomy and unpredictability to them.

That is the lens through which I approach this subject professionally – and the same lens that shapes the world of Cryptogeddon.

The Future Is Leaking Into the Present

AI 2027 turns abstract concerns about automated AI research and recursive improvement into a detailed – and increasingly uncomfortable – scenario.

The most dramatic concern is recursive self-improvement: AI systems helping to design, train or improve their successors, thereby accelerating AI research itself.

The fear is not simply that the next model will be better. It is that AI development could become a feedback loop moving faster than human institutions can understand, govern or interrupt.

That remains a forecast, not an accomplished fact. No frontier laboratory claims to have created a fully autonomous system that can improve itself indefinitely. The timeline is uncertain, and confident predictions about artificial general intelligence have a long history of being wrong.

But uncertainty cuts in both directions.

It does not mean the danger is imaginary. It means we do not know how much time we have – or how reliable our controls will be when capabilities change.

The authors of AI 2027 tried to make that abstract concern concrete. Their scenario imagines AI research becoming increasingly automated, competition intensifying and oversight failing to keep pace.

It is not prophecy, and its authors explicitly say it depicts only one possible future. Its value is that it forces us to examine the chain of events rather than arguing vaguely about whether “superintelligence” will be good or bad.

Parts of that imagined chain no longer feel very distant.

AI Has Already Crossed Boundaries

Anthropic’s research spans alignment, interpretability, societal impacts and frontier testing—including its investigation of Claude models gaining unauthorized access to real systems.

In September, Anthropic published an assessment of four cybersecurity incidents involving different Claude models.

During controlled security evaluations, the systems were supposed to attack fictional targets. A configuration error left them connected to the real internet. The models then gained unauthorized access to real third-party systems.

These were not malicious AIs pursuing some independent plan for world domination.

Anthropic found no evidence that agents coordinated with one another, developed goals beyond their assigned task or tried to evade oversight. The systems were pursuing the objectives humans had given them in an environment whose safeguards had failed.

That is an important distinction.

It is also not especially comforting.

Anthropic concluded that the models displayed biased reasoning and recklessness: they interpreted ambiguous evidence in ways that justified continuing their task, even when their actions could cause real harm. In one case, the same flawed reasoning also persuaded a monitoring system that the activity was simulated.

This is the version of AI risk that concerns me most – not a machine suddenly becoming evil, but a powerful system pursuing a poorly bounded objective with speed, persistence and access that humans cannot match.

From a security perspective, the incident has a painfully familiar shape:
1. A test environment was incorrectly configured.
2. A powerful tool received access it was not supposed to have.
3. The scope of the exercise was insufficiently constrained.
4. The system continued pursuing its assigned objective.
5. Monitoring did not reliably recognize the resulting harm.
Any one of those conditions might have been manageable. Together, they allowed a controlled exercise to reach real systems.

Now add an agent that can operate for hours, write code, discover vulnerabilities and rationalize why it should continue.

No sentience is required.

One Agent Is a Problem. What About Forty-Five?

AI agents can already operate in parallel, coordinate their work and search continuously for vulnerabilities.

Anthropic has also been studying emerging multi-agent systems.

In one experiment, researchers gave 45 agents their own virtual machines and a shared forum, then asked them to collaborate on finding vulnerabilities in open-source software. The swarm found hundreds of vulnerabilities over an extended run.

That can be enormously useful.

Defenders could use the same capability to find and repair weaknesses before criminals exploit them. AI can help security teams examine more code, investigate alerts faster and find relationships that would otherwise be missed.

Attackers can use it too.

The significance is not that AI has spontaneously formed a mob. It is that humans can now assemble persistent, coordinated groups of machine agents at a scale and speed that were previously impractical. Once deployed, those systems may also behave in ways their operators did not anticipate.

Cybersecurity has always been asymmetric: a defender must protect everything, while an attacker needs one opening. AI may magnify that imbalance by making reconnaissance, vulnerability discovery, exploitation and adaptation cheaper and faster.

And the capability is reaching below the software layer.

Anthropic recently reported that an AI system improved known attacks against experimental cryptographic algorithms. The work did not break the encryption protecting today’s banking or email, and it should not be presented as if it did.

But one result improved an attack against a post-quantum signature candidate that had already survived two years of expert review. Another accelerated an attack against a deliberately reduced version of AES.

This is legitimate, valuable research. Finding weaknesses is how cryptography becomes stronger.

It is also a demonstration that increasingly autonomous AI can contribute to highly specialized offensive research once reserved for expert humans.

Capability does not care which side uses it.

The Problem Is Not Just the Model

The system did not escape from a perfect enclosure. The enclosure was never as isolated as people assumed.

One of the recurring mistakes in conversations about AI safety is treating the model as though it exists in isolation.

An AI system becomes much more consequential when it is connected to tools, proprietary data, source-code repositories, cloud infrastructure, identity systems and other agents. Its risk depends not only on what the model knows, but on what the surrounding system allows it to do.

This is where AI security becomes an architectural problem.

Organizations need to understand which models are being used, what data reaches them, which tools they can invoke, how their behaviour is monitored and what happens when they act outside the expected path. Traditional controls still matter, but they must be adapted to systems that reason, generate and act.

That is also why AI safety cannot be reduced to teaching a model to refuse a dangerous question. A refusal is one control. It is not a security architecture.

The more agency we give these systems, the more we need defence in depth: constrained permissions, isolation, testing, monitoring, human approval for consequential actions and the ability to stop behaviour that was not anticipated.

The challenge is that capability is moving quickly while the practices for securing it are still being built.

The People Building It Want the Ability to Slow Down

In July, more than 1,300 employees of frontier AI companies signed Pacing the Frontier, calling for international mechanisms that could deliberately slow automated AI development if the risks demanded it.

The signatories include senior figures from OpenAI, Anthropic, Google DeepMind, Meta and other leading laboratories.

Their argument identifies a brutal incentive problem.

A laboratory may believe that slowing down is prudent while also believing that it cannot afford to let a competitor – or another country – move ahead. Everyone can recognize the danger and still keep racing toward it.

This is why voluntary promises are not enough. It is also why simple calls to “stop AI” are unlikely to work. The technology offers too much economic, scientific and military advantage.

The question is whether governments and companies can build credible evaluation, monitoring, containment and coordination mechanisms before a crisis forces them to improvise.

The International AI Safety Report 2026 gives this concern broader weight. Led by Yoshua Bengio and written by more than 100 experts with an advisory panel nominated by over 30 countries and international organizations, it reviews malicious use, technical failures and systemic risks.

Its conclusion is not that catastrophe is certain. It is that capabilities are advancing quickly, real-world evidence of harm is accumulating and important safeguards remain unproven against sophisticated attacks.

That is the sober version of the warning.

We do not need to accept the most apocalyptic scenario to agree that the safety margin is thin.

So, Will AI Kill Everyone?

WIRED examines why recursive self-improvement, agentic swarms and recent security incidents are alarming researchers inside leading AI labs.

Probably the wrong question.

It encourages us to choose between ridicule and panic. If extinction is not imminent, we tell ourselves everything is fine. If it is possible, we jump directly to Terminator imagery and lose sight of the practical decisions in front of us.

There is an enormous spectrum of harm between a hallucinated answer and the end of humanity.

AI can amplify fraud and disinformation. It can discover and exploit vulnerabilities. It can lower the expertise required to conduct sophisticated attacks. It can make consequential decisions at machine speed. It can behave recklessly when instructions, incentives and safeguards do not align. It can concentrate power in a small number of companies and governments.

None of that requires consciousness, hatred or a secret desire to replace us.

The existential question still matters. A low-probability event that ends civilization deserves attention even if nobody can assign it a trustworthy percentage.

But the strongest reason to take AI safety seriously is not that we know the machines will kill us.

It is that we are building systems with growing autonomy, imperfectly understood behaviour and access to real infrastructure – and placing them inside a global race whose participants believe they cannot afford to pause.

That is frightening enough.

And Yes, It Is Excellent Material

My professional work asks how these systems can be secured. The novelist in me asks what happens when all the controls fail at once.

As someone who has spent decades in cybersecurity, I find these developments alarming. My current work in AI security makes them impossible to dismiss as somebody else’s distant problem.

As the author of Cryptogeddon, I also cannot look away.

Near-future fiction lives in the narrow space between what exists and what might plausibly happen next. That space is shrinking. Ideas I once would have used to make the novel feel speculative now appear in research papers and incident reports before I can finish writing them.

My professional work helps me see how the pieces fit together: not only what a new AI system can do in a demonstration, but what might happen when it meets real networks, imperfect controls, organizational pressure and human fallibility.

Then the novelist in me asks the next question:

What happens when all of those things fail at once?

There is a strange bargain in that. The world keeps making the subject more urgent while making the fiction harder to outrun.

I wish some of this material were less plausible.

But I would be lying if I said it did not make me want to write faster.

Sources and Further Reading
September 19, 2026
Where AI Fits Into My Writing Process – and Where It Doesn’t
I use AI in almost everything I write. That does not mean AI does the writing.

AI can surround the writing process without becoming the writer.

I use artificial intelligence in almost everything I write – but probably not in the way people assume.

That sentence requires some explanation, because “I use AI” has become almost meaningless.

It can mean asking a search engine for help finding a source. It can mean using a chatbot to challenge an argument or organize research. It can also mean entering a prompt, receiving several thousand words and publishing them under your own name.

Those are not the same thing.

So, before I describe my process, I want to draw the most important line clearly:

AI does not write my books. Every word in my manuscripts is my own.

I use AI while researching Cryptogeddon, Holy Wars and other projects. I use it to explore technologies, interrogate ideas, identify gaps in my knowledge and find sources worth reading. But when it is time to write the book, I write it.

My blog workflow is somewhat more flexible. A weekly article has a different purpose, lifespan and production cycle than a novel. AI may help me organize the argument, test structures, identify repetition or refine a difficult passage. Even there, however, I am not asking it to invent a subject and then publishing whatever comes back.

AI participates in the process. It does not own the work.

Here is what that actually looks like.

The Process Begins Before AI

My writing rarely begins with a prompt.

It begins with something that catches my attention: a book, a news story, an historical question, a development in cybersecurity, an experience at the gaming table or an idea that refuses to leave me alone.

I capture those thoughts in Apple Notes. Some are only a sentence. Others grow into collections of links, quotations, questions and fragments. Most never become finished pieces at all.

Before I involve AI, I try to know at least three things:
1. What am I curious about?
2. Why does it matter to me?
3. What question am I trying to answer?
Those decisions need to come from me. Otherwise, I am not using AI to develop an idea. I am asking it to supply one.

That may produce content. It is unlikely to produce anything only I could have written.

Stage One: Mapping the Territory

Once I have a subject, I often use AI to help me understand its shape.

Suppose I am researching a technology for Cryptogeddon. I may begin with broad questions:
- How does this system work?
- What would have to go wrong for it to fail?
- Who controls it?
- What assumptions does it depend upon?
- How might an attacker abuse it?
- What secondary consequences am I overlooking?
- What terminology should I understand before researching further?
The purpose is not to collect prose for the manuscript. It is to improve my mental model of the subject.

AI is particularly good at revealing the structure of an unfamiliar field. It can identify major concepts, show how they relate and suggest questions I would not have known to ask. That makes it a useful starting point.

But a starting point is not a source.

AI helps me map unfamiliar territory, identify connections and discover better questions.

Stage Two: Moving From Answers to Sources

Language models are dangerously good at sounding authoritative.

They can present established fact, reasonable inference and complete invention in exactly the same confident tone. A polished answer can be helpful, but fluency is not evidence.

My research loop therefore looks like this:
1. Ask AI to help map the subject.
2. Identify the claims that matter.
3. Find the original or most authoritative sources.
4. Read those sources myself.
5. Compare competing interpretations.
6. Ask better follow-up questions.
7. Keep only what the evidence supports.
For technology and cybersecurity, that may mean technical documentation, research papers, incident reports or reporting from sources I trust. For Holy Wars, it may mean historical scholarship, primary accounts and competing interpretations of the same event.

The AI conversation helps me navigate. The sources determine what I can responsibly claim.

My rule is simple:

AI can help me discover a claim. It cannot be the authority for that claim.

Stage Three: Using AI to Create Friction

The most valuable thing AI gives me is not an answer. It is resistance.

Once I have developed an idea, I can ask the machine to attack it:
- What is the strongest objection?
- Which assumption is doing too much work?
- What evidence would weaken my conclusion?
- Am I confusing correlation with causation?
- Which stakeholder am I ignoring?
- Is this genuinely plausible, or merely convenient for the story?
- What would an informed critic say?
This is especially useful because writers become attached to their own ideas. After enough time with an argument or story, it becomes difficult to see what we have assumed rather than established.

AI does not eliminate that problem, and its criticism is not automatically correct. But it can create enough distance for me to reconsider something I had begun treating as settled.

Sometimes I reject its objection. Sometimes I return to the research. Occasionally, it exposes a weakness that changes the direction of the work.

The decision remains mine, but it is a better-tested decision.

Stage Four: The Process Splits

This is where my book and blog workflows become different.

For books

Research, questions, timelines and technical explorations may all involve AI. The manuscript does not.

When I move from research into scenes, narration, dialogue and chapters, I write the words myself. That boundary is deliberate.

A novel is more than an efficient delivery system for a plot. Its language carries the author’s sensibility: what receives attention, what remains unsaid, how a character is judged, where a sentence accelerates and where it pauses. Those decisions accumulate into voice.

I do not want to outsource that discovery.

I may later use tools to help locate inconsistencies or interrogate whether some technical element is believable. But I do not ask AI to generate chapters for me, rewrite my prose in bulk or manufacture a voice I can claim as my own.

Every sentence in the book has to pass through my mind and my hands.

For blog posts

A weekly blog operates on a shorter cycle. I am often responding to something timely, developing an argument in public or sharing work in progress.

Here I may use AI more directly as an editorial tool. Depending on the article, I might ask it to:
- compare two possible outlines;
- suggest a clearer order for ideas I have already assembled;
- identify repetition;
- flag an unsupported leap;
- show where a reader might misunderstand me;
- test alternative headlines;
- or help tighten a passage that is not working.
That is closer to an extended editorial conversation than manuscript generation.

The topic, underlying argument, personal perspective and final judgment still have to be mine. I decide what the article says. I decide which suggestions are useful. And I remain responsible for every sentence published under my name.

The distinction is not that books are sacred while blogs do not matter. It is that the tools are allowed closer to the prose in one workflow than in the other – and even there, they remain tools.

Stage Five: Knowing When AI Is Making the Work Worse

AI has a gravitational pull toward competent blandness.

It likes orderly explanations, symmetrical lists and conclusions that neatly restate whatever came before. It can turn an awkward but interesting thought into a polished paragraph that sounds like it could have been written by anyone.

That is useful when clarity is the problem. It is destructive when the awkwardness contains the writer’s voice.

There are warning signs that AI is getting in the way:
- the writing becomes smoother but less specific;
- every argument acquires an artificial balance;
- uncertainty is replaced with a tidy lesson;
- the same phrases and rhythms begin appearing repeatedly;
- the prose explains things the reader already understands;
- or the article sounds finished before I have decided what I think.
That last danger may be the most serious.

Writing is partly how I discover what I believe. If I accept a polished formulation before doing the underlying thinking, I may end up with something coherent that is not actually mine.

The struggle to articulate an idea is not wasted motion. Often, it is where the insight comes from.

Stage Six: The Final Human Pass

Before anything is published, I want to be able to answer several questions:
- Is this accurate?
- Do I believe it?
- Does it sound like me?
- Is there something specific here, or only a competent summary?
- Have I distinguished fact from inference?
- Am I saying anything I would be unwilling to defend?
- Does every sentence earn the right to carry my name?
AI cannot answer those questions for me because they are not simply questions about text quality.

They are questions about authorship.

The final decision to publish is not a mechanical checkpoint at the end of the pipeline. It is the point at which I accept responsibility for the work.

What AI Is – and Is Not – in My Writing

If I reduce the process to its simplest form, AI plays four useful roles:
1. Research guide – helping me map unfamiliar territory and locate questions worth pursuing.
2. Sceptical reader – testing assumptions, arguments and plausibility.
3. Organizational tool – helping me compare structures and manage complicated material.
4. Editorial assistant – identifying repetition, ambiguity and weak transitions, particularly in shorter work.
What it is not is the novelist.

It does not create the pages of Cryptogeddon. It does not choose the language of Holy Wars. It does not decide what my characters fear, what they value or what their choices mean. It does not turn personal experience into something I can falsely claim to have written.

AI can help me arrive at the blank page better informed, more thoroughly challenged and with a clearer sense of what I am trying to accomplish.

Then I have to write.

When the Research Starts Resembling the Fiction

For a near-future writer, the distance between research and fiction is getting uncomfortably small.

Lately, AI has begun playing one more role in my writing process: subject matter.

There is an old expression that reality is stranger than fiction. For someone writing a near-future cyberthriller, that is becoming less an observation than a weekly professional hazard.

I imagine autonomous systems, coordinated cyberattacks and institutions struggling to control technologies they barely understand. Then I return to the research and discover events uncomfortably close to what I had imagined.

It is getting harder to invent an AI future that does not begin arriving before I finish the draft.

That is unsettling.

It is also irresistible material.

Next week, in The Cryptogeddon Briefing, I am going to look at some of those recent developments – and ask a question that no longer sounds quite as melodramatic as it once did:

How scary is AI getting?
September 13, 2026