With cybersecurity spending expected to exceed $300 billion globally this year, organizations are accelerating their investments in artificial intelligence-based security platforms. Security teams and cybersecurity professionals, however, are confronting issues such as the lack of visibility into how employees are deploying AI across the network.
At the same time, the number of AI-based attacks is growing.
Specifically, nearly 48 percent of cybersecurity professionals report that they lack visibility into how employees deploy AI tools across corporate networks, raising fresh concerns about "shadow AI" use within organizations. At the same time, attackers are now using these virtual chatbots and other technologies as part of their arsenal, with approximately 52 percent of cyber pros reporting that AI is helping threat actors more than defenders.
These results are part of a report released by cybersecurity firm Bitdefender, which surveyed 1,200 IT and cybersecurity professionals, including C-suite leaders, middle managers, and security practitioners.
The numbers also show that just a few years ago, AI-based threats were considered mostly theoretical. Now, however, only 17 percent of those surveyed believe that AI attacks are "hyped." Instead, cybersecurity professionals are experiencing numerous threats that use AI technologies, including:
- Fifty-nine percent report their organization has experienced social engineering attacks they believe involve AI.
- Fifty-five percent report that their organization has experienced malware-based attacks that involve AI.
- Seventy percent report that they are seeing more sophisticated phishing attacks powered by AI.
In the report, Bitdefender's own researchers detail how one nation-state threat group, APT36, has used an AI-driven development model. Other security researchers documented how a cybercriminal group utilized AI to graduate from small-scale hacking schemes to full-blown ransomware attacks.
The Bitdefender report stressed that while an "AI-attack apocalypse" has not materialized, the willingness of attackers and cybercriminals to use AI – as well as employees bringing unauthorized or unverified virtual chatbots and tools into the network – can increase the risks that cybersecurity professionals must now consider.
"Adversaries are using AI to operate at a scale and speed that makes traditional, static defenses obsolete. The window between a vulnerability's discovery and its exploitation has shrunk from months to days, and soon it will be merely minutes," said Chandra Gnanasambandam, CTO at SailPoint.
Adding to cybersecurity professionals' concerns are data breaches, whether related to AI misuse or other security factors. The Bitdefender report found that 52 percent of respondents felt pressure to keep a breach confidential following a security incident.
Flushing AI From the Shadows
The issue of shadow AI is one that cybersecurity professionals began to recognize almost as soon as commercial large language models (LLMs), along with virtual chatbots and assistants, became available to the wider public.
The Bitdefender report shows that the problem of shadow AI is growing. Digging into the numbers, researchers found that 52 percent of cyber professionals report they have full visibility into which AI tools their organization and its employees use. In contrast, 45 percent report they are partially blind to AI use within their organizations, while 2.5 percent are completely in the dark.
At the same time, some organizations are rapidly deploying AI without fully understanding what those systems are doing once they're running in production. A recent Cloud Security Alliance report found that 82 percent of organizations lack visibility into AI runtime behavior, highlighting a growing security blind spot, said Crystal Morin, a senior cybersecurity strategist at Sysdig.
AI applications often have access to sensitive data, self-authentication with APIs, and the ability to autonomously trigger downstream workflows, making them attractive targets and powerful pivot points for attackers, Morin added.
"When security teams lack runtime visibility, they lose the ability to tell the difference between an AI model behaving normally and one that's been manipulated or compromised," Morin noted. "By the time they figure it out, the damage is already done and credentials are long gone."
Security teams also need to treat AI-assisted development as part of the attack surface, observed Nicole Carignan, senior vice president for security and AI strategy and field CISO at Darktrace.
In addition to visibility into the code produced, cybersecurity professionals need to understand the tools, functions and integrations that the code relies on.
Organizations should understand which internal and external tools are being invoked, how functions interact and what trust relationships are being created. Graph analysis of tool and function calls—across both internal systems and external services – is essential for identifying unexpected paths, privilege escalation and unsafe data flows.
"AI-generated code often pulls in libraries, APIs or services automatically, sometimes with little transparency to the person building the application," Carignan said. "Human analysis of these dependencies is still required to understand ownership, maintenance, security posture and long-term risk. AI can assist with this process with cyber-AI models that can help identify vulnerabilities, insecure patterns and known weaknesses in generated code, but it should augment, not replace, expert judgment."
Countering AI-Enhanced Threats
One reason cybersecurity professionals are increasingly concerned about AI-enhanced attacks and issues like shadow AI is that the technology has fundamentally changed the risk-acceptance landscape. In years past, security pros could treat a vulnerability as manageable to help ensure the organization met its goals. AI has now changed all that, said Ronald Lewis, head of cybersecurity governance at Black Duck.
"We're seeing the same dynamic play out with AI adoption. Organizations are rewarding speed to deployment, which puts direct pressure on security teams to approve AI solutions quickly," Lewis said. "Those approvals are often based on deterministic evaluation models – treating AI like traditional automation – where outcomes are assumed to be predictable and governed by fixed rules. However, AI doesn't behave that way, and that mismatch is introducing risk under the illusion of control."
Instead of looking outside an organization, cybersecurity teams can counter some of these AI threats by focusing more on network access and identity. This includes enforcing least-privilege access and zero trust principles, while also recognizing that managing non-human identities is fundamentally different from governing humans, said SailPoint's Gnanasambandam.
"The expertise required is less about a specific, narrow skillset and more about a strategic understanding of modern, identity-centric security architecture," Gnanasambandam added. "This expertise is often cultivated internally by upskilling existing security and IT teams to adopt this new, identity-focused paradigm. It can also be found by partnering with security vendors who are building the architectural foundation for real-time governance and agentic security."
The Complexities of Data Breach Reporting
Bitdefender researchers noted that when they first asked IT and security professionals about data breaches in 2023, 42 percent of respondents reported they had been asked to keep a breach confidential. Since then, that number has increased.
The report noted that 2023 marked the year that new U.S. Securities and Exchange Commission (SEC) rules requiring better and faster disclosures from public companies following a breach took effect.
Since then, other laws and regulations have gone into effect in the U.S. as well as globally.
If a breach involving sensitive personal or corporate data is confirmed, organizations are typically required to assess notification obligations under applicable privacy regulations. For example, under the EU's GDPR, companies must notify regulators within 72 hours of becoming aware of a personal data breach, while U.S. state laws impose varying timelines and disclosure requirements depending on the data involved and the residency of affected users.
"Meeting these obligations depends heavily on having visibility into who had access to what data, when, and under which permissions," said Chris Radkowski, a governance, risk and compliance expert with Pathlock. "Without strong identity governance and access controls, organizations often struggle to quickly define the scope of an incident, which can delay notifications and increase regulatory and operational risk."
