ACCORDING to Dark Reading, agentic AI is widely considered the next major target for cybercrime, with nearly half (48%) of respondents expecting it to be the top attack vector by the end of 2026. The piece highlights that organisations are adopting agentic AI to streamline operations and foster productivity, even as experts warn this could expand the attack surface and lead to insecure code if rapid adoption outpaces security.
It also notes that deepfakes are seen by 29% of respondents as a major social engineering vector aimed at “big fish” such as Fortune 500 companies and governments, a concern tied to the broader rise of open‑source AI agents and shadow AI in workplaces. The article quotes analysts from Omdia and others warning that traditional guardrails are insufficient and that continuous authorisation on every resource touched by agents will be essential.
Overall, the piece portrays a security landscape where agentic AI’s growth brings both productivity opportunities and heightened risks that boards and security teams must urgently address, as per the Dark Reading readership poll.