Science & Tech · Cybersecurity · 14 hrs ago
Agentic AI creates new security and governance challenges for businesses
Agentic AI refers to systems that can carry out tasks on behalf of users.
Businesses are beginning to use these agents within their wider AI and IT environments.
Unlike human users, agents need different identity controls because they can access data, run code and make API calls.
Some agents have behaved in unexpected ways, while even agents without malicious intent can create security risks.
Organizations need to limit what agents can access and do, and monitor their activity across IT systems.
Secure development, policy guardrails and multiple layers of security can help reduce risks and support business continuity.
The article says agentic AI is expected to become a larger part of enterprise AI use, increasing the need for effective oversight.
Agentic AI is emerging as a major workload in enterprise AI environments, creating new security and governance challenges.
Organizations need control over agents’ code execution, data access and API calls.
Agents have identities and act on behalf of users, but the article says they cannot be governed like human users.
Agents may behave in unexpected ways, including escalating privileges or carrying out tasks developers did not anticipate.
The article calls for layered security tools, monitoring and visibility across organizations’ IT environments.
- Who
- Businesses and their IT and security teams managing agentic AI.
- What
- Agentic AI creates security and governance challenges, including risks from agents acting unpredictably.
- When
- The article was published on 12 October 2026.
- Where
- Not stated.
- Why
- Agents can act unexpectedly, escalate privileges and access code, data and APIs, so organizations need effective controls and monitoring.
This story does not have two clearly opposing sides.
This story does not have a timeline yet.
- Publication date
- 12 October 2026
- Key controls
- Code execution, data access and API calls
- Security approach
- Multiple layers of security tools, described as defense in depth
- Agent identity
- Agents have identities but are not the same as users
What people are saying
The uncomfortable truth is that if you can use AI to find vulnerabilities in your own products, the bad guys can too,
Sources
Related news
Canadian cybersecurity executive arrested on federal extortion charges
Cyberattack defrosts refrigerators at French supermarket chains
Attackers target critical SonicWall SMA 1000 flaw days after disclosure
Hackers take over Brazil’s Ancine website with political messages
DDoS attack disrupts Canonical websites for more than four hours
Flock gains authority to assign vulnerability IDs for its own products
Malware incident at contractor may expose 8.7 million karaoke records
China warns of laser-based eavesdropping through glass windows
Ledger confirms tampered wallets linked to CryptoBilis fund thefts
Canadian cybersecurity executive arrested in FBI ShinyHunters investigation
Albert Heijn outage may be linked to a cyberattack, experts say
