Security in the Age of AI-Powered Cyberattacks

Artificial intelligence is transforming virtually every facet of the modern world, introducing new ways to work, communicate and live. In the realm of cybersecurity, however, one of its most consequential effects isn’t the creation of new threats, but rather, the acceleration of familiar ones.

Phishing, malware and other longstanding schemes are becoming faster, more convincing and easier to deploy at scale, according to Comcast Business’s 2026 Cybersecurity Threat Report. Bad actors have more tools at their disposal than ever, posing a major threat to businesses, whose operations risk disruption and whose customers may lose trust. Comcast Business detected more than 79.3 billion cybersecurity events targeting its security customers over 12 months, including 25 billion phishing events and 22 billion drive-by compromise events. 

“Attacks are very similar to what they have been in the past, but they’re happening at machine speed now,” says Amit Verma, Chief Technology Officer of Comcast Business. “AI is also expanding the volume of the attacks.”

Modern phishing is emblematic of how AI can amplify a familiar threat. Fraudulent messages that once contained obvious errors or awkward phrasing can now be generated articulately and tailored to individual targets. And social engineering, a key tactic used in phishing, is easier when AI can create convincing deepfake voices. 

On top of that, threat actors have a lower barrier for entry on a technical level. Attackers can use AI to help develop malware and ransomware and better search for vulnerabilities. Comcast Business observed nearly 289 million scanning events over the past year. 

Instead of limiting such reconnaissance to a specific organization or set of public IP addresses, attackers can continuously scan devices and systems, catalog potential vulnerabilities and exploit them whenever an opportunity arises. They can also find ways to combine vulnerabilities that might otherwise pose low risk on their own, Verma says, chaining them together to find an exploit. 

The sheer scale of these attacks works in favor of threat actors. With billions of phishing attempts, even a tiny percentage succeeding can pose a significant threat to enterprise.

But the same capabilities that are accelerating cyberattacks can also give defenders new tools to identify and stop threats. Advanced AI models can help security teams search for vulnerabilities at a speed and scale that would be difficult to do manually, allowing businesses to discover and address critical weaknesses before attackers can exploit them.

The risks aren’t limited to the ways that threat actors themselves are using AI. As companies integrate AI into their own operations, they’re also creating new points of exposure. 

One potential vulnerability is “shadow AI,” or the unauthorized AI applications often used by employees. Staffers using tools that haven’t been vetted by their organizations might inadvertently share proprietary or sensitive information with outside applications.

The emergence and rising prominence of agentic AI raises additional questions. AI agents often require access to multiple systems and data sources to perform their work, creating a growing number of non-human identities in corporate networks. 

“If AI agent credentials are compromised, the attacker could gain access to much more data and many more systems than a traditional application,” Verma says. Because agents can be authorized to take actions, attackers that gain control of their credentials could alter workflows while appearing like legitimate users — making identity management and data loss prevention a crucial part of enterprise cybersecurity. 

For executives eager to reap the benefits of AI, Verma says, the goal should be to manage those risks without impeding adoption. “You cannot throttle AI innovation because you are very security focused, but you cannot just adopt any AI tool without investigating further,” he says. That can mean creating processes to assess AI tools before they’re deployed — including the data they can access and the actions they can take — as well as monitoring their behavior after deployment.

Strategies used by businesses have often focused on perimeter security, such as firewalls. While these defenses are still important, their efficacy becomes limited once an attacker has compromised a device or agent and gained access to the network. The consequences of such a breach can be substantial, including theft of data, ransom or operations disruption, potentially affecting revenue and customer trust. A serious cyberattack can also slow innovation if company leaders respond by imposing overly rigid controls on the adoption of new technologies. 

That sensitive balance means it’s critical for businesses to know how to bounce back — to be able to identify, mitigate and recover from incidents rather than prevent every attack. Prioritizing several factors can be key, Verma says: identity management, end-to-end observability and strong AI governance. Organizations can control the permissions given to AI agents, maintain records of the actions they take and require human approval for workflows with higher risk. Training employees to identify increasingly sophisticated phishing and voice-based attacks can add another layer of security. 

“Cybersecurity is not just an IT risk anymore,” he says. “It’s a business resilience risk.”

Get smart on today’s evolving threat landscape with the 2026 Comcast Business Cybersecurity Threat Report.