Cyera’s Assaf Morag argues that AI-assisted cybercrime makes one defensive question more urgent: what sensitive data can a compromised account or token actually reach? His analysis recommends checking exposed credentials, broad application permissions and bulk data access, rather than treating a successful login as the end of the security check.
Cyera Watch analysis
What happened
Morag’s analysis examines how AI can support different stages of data theft: prioritising exposed credentials, helping attackers work through cloud environments, analysing stolen files and automating parts of voice phishing.
The article draws on accounts attributed to Anthropic, Google Threat Intelligence Group and EclecticIQ. It is a synthesis of reported activity, not a newly disclosed Cyera experiment or a measured comparison of AI-assisted and conventional attacks.
Its practical recommendations are concrete. Identify exposed credentials and API keys; review OAuth applications and integrations with broad access; map which people and machine identities can reach sensitive information; and check whether unusual bulk access or exports would trigger a prompt response.
Why it matters
An attacker using a legitimate session may not need to defeat the login system again. The permissions attached to that session can determine how much data is available, including through ordinary application interfaces and export functions.
Morag’s argument is that AI can reduce the manual effort needed to explore and exploit that access. The useful defensive consequence is not to abandon authentication controls. It is to connect them to a clear picture of reachable data and a response process for suspicious use.
Our read
The strongest takeaway is an audit question, not an AI panic button: if this identity were compromised, what could it collect before anyone intervened?
Start with an important application or integration and trace its access to sensitive data. Then test whether your monitoring would notice an unusual collection or export. A green login tick is reassuring, but it is not a character reference.
Keep the evidence distinction intact. Morag’s cited incidents illustrate his argument; they do not establish that every attacker has the same capabilities or that every intrusion now follows the same timetable.
What to watch
- Whether organisations can connect identity permissions to specific sensitive datasets, rather than maintaining separate inventories.
- Whether bulk-access alerts produce timely investigation and containment, not simply another dashboard entry.
- Whether further threat research distinguishes AI’s contribution from the automation attackers already used.
Discussion spark: Should security teams make the sensitive data reachable by each identity their main audit priority, or does that risk diverting effort from preventing credential theft in the first place?
Sources and evidence
- ShinyHunters at Scale: A Case Study in AI-Assisted Cybercrime (1 October 2026, 15:21 UTC)
not affiliated with or endorsed by Cyera