AI-enabled breaches rose 56% in a year. The Threat Landscape Series finale covers how AI is accelerating attacks, creating new attack surface, and what to prioritize for 2027.
Part of InfoSight's Threat Landscape Series. This is the final installment.
Over this series we've covered rogue access points and evil-twin Wi-Fi attacks, social engineering that no longer reads like a scam, ransomware operators using AI coding tools mid-intrusion, an Azure directory leak that exposed service accounts and Global Admin names, and the blind spots that let attacks slip past a SIEM.
They look like separate stories, but they share a common thread: AI is making each of them faster, cheaper and harder to spot.
AI is creating new attack surface that most security programs aren't yet built to see.
IBM's 2026 Cost of a Data Breach Report puts a number on it: one in four malicious breaches was AI-enabled, a 56% increase over last year. Those breaches cost an average of $6 million, about $1 million more than the record global average of $4.99 million (IBM).
So the AI security boom isn't hype. It has two sides: attackers using AI against you, and the AI your own organization is adopting faster than it can govern.
Trend 1: AI Is Lowering the Cost of Attacking You
The biggest change isn't that attackers have new superweapons. It's that offensive skill has become cheap. In a contributed piece for The Hacker News, offensive security firm BreachLock described "vibe hacking": using natural-language AI prompts to turn intent into working attack activity. Its conclusion: "The barrier to entry is dropping fast"
That shows up in three ways we covered earlier in this series:
The practical takeaway: an annual pentest and a yearly awareness training module were built for a slower adversary. Validation now has to be more frequent.
Trend 2: Your Own AI Is Now Attack Surface
The second half of the boom is inside your walls. Employees, developers and business units are adopting AI faster than security teams can review it, and much of it arrives without approval.
IBM found that more than 20% of organizations reported a breach targeting AI models or applications. Of the organizations that reported AI-related breaches, 92% lacked proper AI access controls, and only 40% of organizations use access controls on AI models and data at all.
That risk takes three forms:
Trend 3: The Blind Spots Haven't Gone Away, and AI Exploits Them Faster
AI doesn't need new weaknesses when the old ones still work. Each blind spot this series covered gets more dangerous when an attacker can move at machine speed:
The defensive side of the boom matters here. IBM found that organizations using AI and automation extensively in security operations cut breach costs by nearly $2 million on average, yet one in four organizations still haven't adopted these tools (IBM). AI is a force multiplier for whichever side uses it well.
What to Prioritize Going Into 2027
Most organizations can't fix everything at once. This is the order we'd tackle it in:
How InfoSight Helps
InfoSight's approach is simple: Powered by AI, Fueled by Experts. Our certified practitioners use AI to move faster, and they validate every finding by hand. Results are tracked over time in our Mitigator® platform, so you can show leadership and examiners that risk is going down.
| The risk | Where to start |
| Shadow AI, OAuth-connected apps, Copilot data exposure | AI Governance & Exposure Assessment: AI app discovery, OAuth consent review, Copilot and SharePoint exposure analysis, and a 30-90 day roadmap |
| Chatbots, copilots, RAG pipelines and agents | AI Penetration Testing: prompt injection, data leakage and excessive agency testing, mapped to the OWASP Top 10 for LLM Applications, NIST AI RMF and MITRE ATLAS |
| Identity, privileged access and non-human identities | AD & Entra ID IAM Security Assessment and M365 Security Assessment |
| AI-crafted phishing and pretexting | Social Engineering & Physical Security Testing and Employee Security Awareness Training |
| Faster, more frequent attacks | Penetration Testing and Red, Blue & Purple Team Testing |
| Detection gaps between assessments | 24x7 MDR & SOCaaS and Managed SIEM & XDR |
| No AI policy or governance owner | Virtual Security & GRC Leadership (VSGRC) |
Not sure where AI is already touching your data? Most organizations find more than they expect. Request an AI Governance & Exposure Assessment to see which AI tools are connected to your environment, what they can reach, and what to fix first.
Thanks for following the Threat Landscape Series. Catch up on earlier installments on the InfoSight blog.
Subscribe to our newsletter to keep you updated on the latest cybersecurity insights & resources.
One follow-up from a security expert—no spam, ever.
Enter your details below to download the PDF.