Threat actors stories
Many firms cannot see where their AI agents are, leaving identity, policy and supply-chain risks to grow as deployments scale.
Businesses face rising exposure as AI is used to sharpen phishing, while insecure in-house tools and weak controls widen attack surfaces.
Leak-site noise is making it harder for firms to tell real breaches from extortion theatre, as active sites hit 91 in the first quarter of 2026.
Attackers could soon exploit software flaws faster and at scale, as security firms say AI is narrowing defenders' response time.
Illicit discussions of AI tools surged 1,500% in late 2025 as attackers used them to speed up vulnerability hunting and exploitation.
As cyber tools become more powerful, Anthropic is limiting access while OpenAI is widening it, raising fresh fears over misuse.
Businesses facing faster AI-driven cyberattacks will get new Google Cloud tools to spot threats, block fraud and secure agents across workloads.
Most respondents still trust consumer chat apps for sensitive work, despite widespread confusion over what encryption does not protect.
Shorter attack windows are pushing cloud teams towards automated defence, as Sysdig says AI-driven threats now outpace manual response.
It aims to replace fragmented feeds by combining risk scoring and context on millions of IPs and domains for security teams.
New extortion-only gangs are reshaping a ransomware market that remained at about 150 to 200 victim posts a week in the first quarter.
Seven critical weaknesses were found in live production systems over a weekend, showing AI-driven pentests can now uncover basic flaws cheaply.
Defenders face faster, harder-to-stop attacks as SANS says AI is now built into phishing, malware and reconnaissance at scale.
More firms are turning identity security budgets to attack path tools as hybrid and AI-heavy environments expose gaps in remediation.
Companies seeking Cyber Essentials certification must now use multi-factor authentication and managed devices, as remote working rules tighten.
Security chiefs say unauthorised access to Anthropic AI's Mythos model shows generative tools could speed phishing, scanning and exploit discovery.
Greater reporting by English councils has pushed logged breaches up 53% in five years, with serious referrals to the ICO also rising.
Hospitals are paying up to avoid costly downtime, as criminals exploit known flaws and buy access for as little as USD $2,000.
Unprepared firms can face NZ$173,000 in ransomware costs alone, with lost clients and trust pushing the real bill much higher.
Ransomware pressure on US firms is intensifying debate over whether broader AI hacking tools will help defenders or aid criminals.