Sublime projects a 33,000% rise in calendar-invite phishing between May and September

Sublime's figure is a projected growth rate. Its post says "the increase from May to September is projected to be ~33,000%" ^[2]. Morning Brew reported the same figure ^[1]. The post puts the first half of September at 1,426% over the full month of August ^[3], and the whole of September, projected, at 2,852% over August ^[4]. The projected September figure is twice the half-month one ^[14]. A rise of about 33,000% is a projected multiple of roughly 331 times the starting level ^[13]. Morning Brew did not report the starting count, so the percentage cannot be turned into a volume of attacks.

The delivery route is simple. Calendar invites carry malicious links past spam filters and need no attachment ^[5]. The link leads to a credential form, a malware download, or a meeting with someone posing as a trusted contact ^[6]. Sublime says one way the invites reach calendars is through AI agents accepting them on the worker's behalf ^[7]. An agent that accepts by default puts the lure on the calendar without a person deciding to. Consumer Affairs suggests checking settings so that not every invite is added automatically ^[8].

The payroll side is a vendor's account of its own product. Workday announced its Trusted Agent Platform at Workday Rising and said every core capability is generally available ^[9]. It says AI agents are executing workflows across HR and finance without anyone logging in ^[10]. It also says every action through its platform stays bound to Workday's permissions, approvals and audit trails ^[11], and that its guardrails block hidden prompt injections ^[12].

Workday's post describes controls on its own platform. Sublime's figure describes calendar invites. An agent accepting an invite and an agent running payroll hold different grants of authority, and in our view the evidence ties an attack only to the first.