Follow us on Google News
AI-powered cyberattacks are accelerating, prompting cybersecurity firms to deploy decoys and automated defensive tools.
How Are Decoys Being Used Against AI Hackers?
Cybersecurity companies are increasingly using deception techniques to counter artificial intelligence-powered attacks, as automated hacking tools reduce the time required to identify and exploit vulnerabilities.
Horizon3 chief executive Snehal Antani said AI models trained on publicly available cybersecurity information can be particularly vulnerable to deliberately planted decoy files. These may include fake login credentials designed to attract attackers and trigger security alerts before they reach critical systems.
Research published by Horizon3 in June found that 21 AI models, including models developed by OpenAI and Anthropic, fell for decoys in 78.5% of tests, compared with 37% for human hackers. Some advanced models were deceived more than 90% of the time.
However, Antani cautioned that sophisticated attackers could recognise such traps. Organisations must therefore maintain other protections, including security patches and restrictions on system access.
https://the420.in/fcrf-certified-chief-ai-officer-second-cohort-november-2026/
How Quickly Can AI Carry Out Cyberattacks?
Research by cybersecurity firm Ensign InfoSecurity found that 10 frontier AI models successfully breached a test system during initial assessments. However, their performance declined during later stages, particularly when attempting to steal credentials and move between systems.
Meanwhile, Palo Alto Networks disclosed that an attacker using frontier AI breached an enterprise network autonomously in under 10 hours on September 2. The company estimated that a comparable attack would have taken several human attackers approximately two weeks. Microsoft also reported in October that AI could reduce the time needed to steal data, discover credentials and infiltrate computer systems from days to minutes.
Fortinet’s Singapore and Brunei country head Jess Ng warned that organisations now have significantly less time to identify and address vulnerabilities before criminals exploit them.
What Threats Are Emerging From AI-Assisted Crime?
AI is also lowering the technical expertise required to conduct cybercrime. Attackers can use automated tools to interpret vulnerability information and develop phishing operations.
One example involved EvilTokens, a platform that enabled criminals with limited programming knowledge to offer phishing-related services. Within six months, it had affected 10,000 organisations across 120 countries before Microsoft and its partners disrupted it in September. Security researchers have also warned that AI agents can combine multiple lower-severity vulnerabilities into more serious attacks.
Recent critical PHP vulnerabilities illustrated the pressure facing defenders. Following their disclosure on October 21, Cloudflare observed hacking attempts reaching millions, although AI involvement in those attempts remained unconfirmed.
How Are Cybersecurity Firms Strengthening Defences?
Cybersecurity organisations are responding by developing AI systems capable of identifying weaknesses before attackers exploit them.
Horizon3 uses AI to conduct authorised testing of organisational networks, identify vulnerabilities and verify whether security fixes are effective.
The company announced in August that it had raised US$250 million, including investment from Singapore’s Economic Development Board, to support expansion and cybersecurity development.
Industry representatives have also called for greater automation and proactive defence. Association of Information Security Professionals president Emil Tan said organisations should anticipate possible attacks rather than respond only after breaches occur.
The growing use of AI on both sides of cybersecurity highlights the need for faster vulnerability management, effective deception systems and stronger automated protections as attacks become increasingly difficult to contain.
