AI reduces the cost of cybercrime and increases risk for all businesses

Welcome to Eye on AI with AI reporter Sharon Goldman. This time… How AI is making cyberattacks cheaper for hackers… US lawmaker says Nvidia helped Deep Seek refine AI model later used by China’s military… Dow Chemical to cut 4,500 jobs in AI overhaul… Inside Anthropic’s plan to scan and destroy millions of books.

One of the things I’m always interested in when it comes to AI is how it impacts cybersecurity. 2 months ago Focus on AIHe quoted a security leader who described the current moment as “tough” as companies struggle to ensure the security of their systems in a world where AI agents no longer just answer questions but act autonomously.

This week I spoke with Gal Nagli, head of threat exposure at $32 billion cloud security startup Wiz, and Omer Nevo, co-founder and CTO of Irregulator, a Sequoia-backed AI security lab that works with OpenAI, Anthropic, and Google DeepMind. Wiz and Irregular recently completed a joint study on the true economics of AI-powered cyberattacks.

Providing AI-powered cyber attacks at bargain prices

They realized that AI-powered hacking has become incredibly cheap. In testing, AI agents completed advanced offensive security challenges at an LLM cost of less than $50. This is a task that typically costs close to $100,000 when performed by human researchers paid to find flaws before criminals do. In controlled scenarios with clear targets, the agent resolved 9 out of 10 attacks modeled on the real world. This shows that much of the offensive security work is already being automated quickly and cheaply.

“It was really surprising, even for a lot of experienced professionals who have looked at both AI and cybersecurity, that we didn’t think that what AI couldn’t do, that models would be able to do,” Nevo said, adding that even in just the past few months, there have been huge leaps in capabilities. One area is AI models that can stay on track to perform multi-step tasks without losing focus or giving up. “Increasingly, models can solve truly expert-level challenges, even for offensive cybersecurity experts,” he said.

This is especially an issue right now, as in many organizations, non-technical professionals such as marketing and design are using accessible coding tools like Anthropic’s Claude Code and OpenAI’s Codex to bring applications to life. These people are not engineers, Nagri explained. “They don’t know anything about security, they just develop new applications on their own and use sensitive data that is exposed on the public internet and is very easily exploited,” he said. “This creates a huge attack surface.”

Cost is no longer an issue for hackers

This research suggests that the cat-and-mouse game of cybersecurity is no longer constrained by cost. Criminals no longer have to choose their targets carefully when AI agents can probe and exploit systems for just a few dollars. In this new economic environment, every system exposed is worth testing. Every weakness is worth trying.

In more realistic real-world situations, the researchers saw performance decrease and costs double. But the bigger accomplishment remains. That means attacks are becoming cheaper and faster to launch. And most companies still defend themselves as if serious attacks require expensive human labor.

“If AI is able to carry out sophisticated attacks, and they can do so at scale, all of a sudden more people are exposed to attacks. [even at] “In smaller organizations, people will need to be much more aware of cybersecurity than they are today,” Nebo said.

That also means using AI in defense will become a critical need, he said, raising the question: “Are we helping defenders leverage AI enough to catch up with what attackers are already doing?”

So, here’s more AI news for you.

sharon goldman
sharon.goldman@fortune.com
@SharonGoldman

The fate of AI

‘New politics of power’: Utilities seek $31 billion in hikes as voters rebel against soaring electricity prices – Written by Jordan Blum

How Samsung’s first-ever chief design officer is reinventing the electronics giant for the AI era – Written by Nicholas Gordon

Mark Zuckerberg predicts “massive acceleration of AI” in 2026, with up to $135 billion in capital investment expected to match, Meta outperforms fourth-quarter revenue – Written by Sharon Goldman

Wabi raises up to $1 billion, partners with Uber to deploy 25,000 robotaxis as competition for autonomous driving supremacy intensifies – Written by Jeremy Kahn

SAP boss Christian Klein looks to the future of AI: What you say will matter more than what you type – Written by Kamal Ahmed

AI in news

The US lawmaker said Nvidia helped DeepSeek hone an AI model that was later used by the Chinese military. A U.S. lawmaker says Nvidia has provided technical assistance to Chinese AI startup DeepSeek, allowing DeepSeek to improve the efficiency of its models and train them in far less GPU time than typical U.S. frontier models, which are later used by the Chinese military, raising new concerns about the transfer of AI technology to Beijing, according to a letter seen by Reuters. Rep. John Moolener, chairman of the House Select Committee on China, cited internal Nvidia documents showing that engineers helped optimize algorithms and hardware, and argued that the episode highlights the need for stricter export controls and enforcement to prevent U.S. AI technology from being diverted to military use by potential adversaries. Nvidia countered that it is unreasonable to think that China’s military is dependent on U.S. technology, and the Commerce Department and DeepSeek did not comment.

Dow Chemical to cut 4,500 employees due to AI overhaul. of wall street journal Dow Chemical Co. has announced that it will cut 4,500 jobs as part of a sweeping cost-cutting initiative that relies heavily on AI and automation to boost productivity and profits as it deals with a widening quarterly loss due to lower revenue and higher costs. The “Transforming to Outperform” program is expected to generate an additional $2 billion in operating income and will incur one-time costs of $1.1 billion to $1.5 billion, including up to $800 million in severance payments. CEO Jim Fitterling said the plan is a “comprehensive and fundamental simplification” of Dow’s operating model. The move comes amid a further wave of layoffs at companies ranging from Amazon to UPS as companies shift spending to AI and technology.

Inside an AI startup’s plan to scan and destroy millions of books. This fascinating story washington post Details in a newly unsealed court filing reveal that Anthropic quietly launched an internal effort called “Project Panama” in early 2024 aimed at “destructive scanning.”[ning] The documents, released as part of a copyright lawsuit that was settled for $1.5 billion, provide valuable information about how aggressively AI companies have pursued high-quality data, especially books, which executives believed would force models to “write well” rather than imitate “low-quality Internet speech.” The filings suggest that Anthropic, Meta, and other major research institutions considered large-scale licensing impractical and instead attempted to gain bulk access, sometimes without the knowledge of authors, by downloading pirated collections such as LibGen, despite internal warnings about legality. The revelations, along with other copyright cases, highlight how the modern AI race has been fueled by a frenzied, often secret race to capture humanity’s written records.

Pay attention to AI numbers

$211 billion

According to the new 2026 AI Funding Report from HumanX and Crunchbase, this is the amount of venture capital flowing into AI in 2025, accounting for 50% of all venture capital globally.

Other important points from the report:

Companies with at least one female founder secured 47% ($84.7 billion) of all AI funding in North America and Europe.
77% of total AI funding, or $163 billion, came from rounds of $100 million or more.
59% of AI-focused investments went into supporting ecosystems such as infrastructure (19%), deep tech/robotics (11%), and AI-powered software for health and security (15%).