StrongestLayer has launched two email-security products: one holds risky payment and account requests for verification, while the other checks messages for hidden instructions before they reach AI assistants. The practical aim is to catch attacks that look ordinary to conventional email filters, including attempts to redirect payments and prompt-injection messages.
Watch Desk analysis
What happened
SiliconANGLE reports that Intent Workflows can hold payment, payroll and executive-account requests until a customer-defined verification step is completed. For payments, the system compares a request with the payee’s history. If a payment or vendor is new, the company says staff must verify it using a phone number already on file; the email is not released until the result is logged. Customers can set their own thresholds and route requests to a security team or process owner.
The second product, AI Instruction Intercept, checks incoming messages for concealed instructions before an assistant such as Microsoft 365 Copilot or Google Gemini processes them. Suspicious messages are quarantined by default, with an explanation of the instructions detected. Both products are available now for Microsoft 365 and Google Workspace and are designed to run alongside existing email security.
Why it matters
The two tools address different ways an inbox can become a route to trouble: a plausible-looking request can prompt an irreversible payment, or a message can try to steer an AI assistant. StrongestLayer says that, in its review of 32,052 attacks that got past existing defences at 45 organisations, payment or bank-change requests made up only 3.6% of the attacks but accounted for more than 98% of estimated financial exposure under its risk model. Those are company-reported figures, not a measure of risk across all organisations.
The useful distinction is between deciding whether an email looks suspicious and checking what it is asking someone or something to do. A familiar supplier’s account may be compromised, and a message need not contain malware or a suspicious link to ask for new bank details.
Our read
Holding the requests that are hardest to reverse, then verifying changes through a known contact route, is a sensible bit of friction in exactly the right place. Intercepting hidden instructions before an AI assistant reads a message is also a concrete defence against inbox-based prompt injection. These are product claims, though: security teams will want to test detection and verification flows in their own environments before treating a new layer as a force field. Email has had enough exciting surprises.
What to watch
- How customers configure holds and verification for payments, payroll and account recovery.
- Whether independent customer results show fewer successful fraudulent requests.
- How AI Instruction Intercept performs across different assistants and types of hidden instruction.
Discussion spark: Should organisations automatically hold every payment or bank-detail change for verification, or is that too much friction for routine business?
Sources and evidence
- Exclusive: StrongestLayer launches tools to verify payment requests and block hidden AI instructions – SiliconANGLE (5 November 2026, 14:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.