Abnormal AI review 2026: features, pricing, pros and cons
Abnormal AI is a behavioural email security tool that catches BEC and fraud in Microsoft 365 and Google Workspace. Here's what it does, what it costs and who should skip it.

Our verdict on Abnormal AI
Abnormal AI is one of the strongest tools for stopping business email compromise and fraud in Microsoft 365 or Google Workspace. Mid-size and large cloud-email companies should ask for a proof of value; small businesses and on-premise shops should probably skip it.
Pricing Quote only; no free plan or trial. Buyer data suggests about $20-35 per mailbox per year for email security plus a $5,000-15,000 platform fee
Visit Abnormal AIAbnormal AI is an enterprise email security company, and if you run Microsoft 365 or Google Workspace for a few hundred people or more, I think it's one of the strongest options for catching the phishing and invoice fraud your gateway lets through. It isn't a consumer app. There's no free plan and no public price list, so you'll be talking to a sales rep before you see a number.
The pitch is easy to follow: instead of scanning for known bad links and attachments, Abnormal learns what normal email looks like for each person and vendor in your company, then pulls out the messages that don't fit. Users who've written reviews mostly like it. The complaints are about price, a few misfiled emails and gaps around the edges, which I'll get to.
Abnormal AI in brief
Abnormal was founded in 2018 by Evan Reiser, who's still CEO. On its about page, the company explains that its founders spent the 2010s building adtech models that predicted human behaviour, then pointed the same idea at cybercrime. For years it went by Abnormal Security; it now trades as Abnormal AI, and older review sites still use the old name (which is why you'll see both in search results).
It's big. Abnormal says it protects more than 5,000 organisations, including 30% of the Fortune 500, a figure it gave in a September 24 press release.
The product line now has three parts. Email Security is the original business and still the reason most people search for it. Identity Security watches sign-ins, SaaS apps and help desk requests for signs that an account's been taken over. AI Security, the newest part, tries to track which AI tools and agents your staff use and what they do. All three run on the same behavioural model, which the company calls Attune 1.0.
Setup is the part that sets it apart from older tools. Abnormal connects to Microsoft 365 or Google Workspace through their APIs, so you don't reroute your mail (no MX record changes). The vendor's line is "Deploy in 60 seconds via API." I'd treat that as marketing shorthand, since tuning and policy work take longer, but going by what reviewers say, getting connected really is quick.
Key features
Behavioural detection for email
The core product, Inbound Email Security, builds a baseline for every employee and vendor and scores each message against it. That's why it's aimed at business email compromise (BEC), vendor invoice fraud and payload-free scams: emails with no dodgy link or attachment for a filter to catch, just a believable request to change bank details. Abnormal claims it finds more than 1,200 attacks per 1,000 mailboxes each month that got past upstream gateways, based on its own June 2026 data. That's a vendor number, so take it as a sales claim rather than an audit.
Automatic removal and calendar invites
When something's flagged, Abnormal pulls it out of the inbox, ideally before anyone clicks. It does the same with malicious calendar invites, which is a trick I don't see many vendors talk about. The catch with any API-based tool is timing: it acts after Microsoft or Google has delivered the message, and one PeerSpot reviewer described forwarded mail reaching a third-party ticketing system before Abnormal could clean it up.
Detection 360 and the threat log
Detection 360 lets your analysts submit misses and false positives, and each one becomes a tracked investigation. One IT manager on PeerSpot said a human expert reviews a submission "Within a few hours". The threat log shows every message Abnormal has judged, with the signals that fired, so you can see why something was blocked instead of guessing.
Custom AI models and rules
You can describe an organisation-specific pattern in everyday language and have Abnormal build a model for it, without writing rules. If you'd rather write rules, there are custom rules across 50+ message attributes. Honestly, I like that it offers both: behavioural AI is great until you need to enforce one very specific policy by Friday.
Identity and AI security add-ons
In July, Abnormal launched Identity Threat Protection, AI Governance and Infiltration Prevention, plus an "AI App Store" for switching products on in one click. In September came AI Employee Guardrails, AI Agent Security and an AI Security Workbench. Only AI Governance was generally available at that launch; the rest were in private preview, with general availability planned for Q4 2026. So if the AI Security suite is why you're interested, check what's actually shipping before you sign.
Integrations
Beyond Microsoft 365 and Google Workspace, the platform page lists connections to Okta, CrowdStrike, Auth0, Ping Identity, Workday, Salesforce and ServiceNow, and it feeds SIEM and SOAR tools. Not bad for a company that started as an inbox filter.
Abnormal AI pricing and plans
There's no pricing page. Abnormal sells through quotes, and you can also buy it through AWS Marketplace if your company prefers to spend committed AWS budget. Capterra's listing marks it as having no free version and no free trial.
What you can get is a proof of value. On the email product page, Abnormal says every evaluation produces "head-to-head miss data against the incumbent", meaning it runs alongside your current tool and shows what that tool missed. For a security buyer, that's more useful than a free tier would be.
For rough numbers, I had to lean on published buyer deal data, so treat these as estimates and not a quote:
| Item | Typical cost (estimate) | Notes |
|---|---|---|
| Inbound Email Security | About $20–35 per mailbox per year at list | Lower per-mailbox rates as seat counts grow |
| Platform fee | About $5,000–15,000 per year | Flat, organisation-wide |
| Account takeover protection | About $5–10 per user per year | Add-on module |
| Example: 1,000 mailboxes | Around $20,000 per year | Median negotiated deal in that data |
| Free plan or trial | None | Proof of value instead |
Is it expensive? Probably, next to what Microsoft already bundles into E5. But it depends on what it replaces. One PeerSpot reviewer said Barracuda quoted them around $170,000 for the features they wanted, while Abnormal came in "just over $60,000". On the other side, an education CISO on the same site said pricing remains a challenge for academic institutions. My guess is it's easiest to justify when it lets you drop a separate secure email gateway, and hardest when it's a pure add-on.
Performance and independent testing
There's no independent lab test of Abnormal's detection rates that I could find and trust, so performance evidence comes from two places: the vendor's own numbers and what customers say.
The vendor numbers first. Abnormal says Attune 1.0 was trained on behavioural signals collected between August and December 2025, and its platform page claims fewer false positives and more unique attacks caught than its previous model. Those comparisons are against its own older model, not against rivals. It also calls itself a Leader in Gartner's Magic Quadrant for Email Security Platforms in its AWS Marketplace listing.
Customers are kinder than I expected.
PeerSpot users give it an average of 9.4 out of 10, and the recent G2 reviews mirrored on AWS Marketplace are mostly four and five stars. A vice president of information security at Comfort Systems USA wrote on PeerSpot that phishing and BEC investigations dropped "from hours to mere minutes" after rollout, because one search finds who got a message, who clicked and where it sits.
The weak spots are real, though. A G2 reviewer in November wrote: "I wish Abnormal AI would catch mail bombing." Another said they still find "quite a few legitimate emails" in their promotions folder. A third admitted they'd seen "quite a few false positives and false negatives" since day one, though it was improving. That's normal for any email filter, I'd say, but it means you'll still want someone checking quarantine.
A couple of structural gaps come up too. One PeerSpot reviewer called the lack of support for on-premise email systems "The biggest pain point", and another wished for outbound scanning. A G2 reviewer noted there's no mobile app and said the interface "isn't great". None of these are dealbreakers for a cloud-first company. They might be for a hybrid one.
Best suited to
Abnormal makes the most sense for mid-size and large companies on Microsoft 365 or Google Workspace that keep getting hit by convincing fraud emails: fake supplier invoices, CEO impersonation, compromised partner accounts. If your finance team has ever paid a fake invoice, this is the kind of tool built for that problem.
It's also a fit if your security team is small and drowning in user-reported phishing, since auto-removal and the abuse mailbox features cut a lot of manual triage.
One PeerSpot reviewer runs it on top of a Proofpoint gateway, while two others moved over from Mimecast.
I wouldn't recommend it to a small business with 20 mailboxes. The platform fee alone probably makes the maths ugly, and Microsoft's or Google's built-in protection, plus some security awareness training, is likely enough at that size. It's also a poor fit if you still run on-premise Exchange.
Abnormal AI alternatives worth considering
Proofpoint is the big incumbent. Pick it if you want one vendor for email security, data loss prevention and insider-threat tools under a single console, and you don't mind a heavier setup. It's also the gateway one of Abnormal's own PeerSpot reviewers kept in front of it, so the two often end up working side by side.
Mimecast has been protecting business email for more than two decades and says it's trusted by over 42,000 organisations. It's worth a look if you want a traditional gateway plus archiving and collaboration protection from one supplier.
Microsoft Defender for Office 365 is the one to price first if you're already on Microsoft 365 E5, since Plan 2 is included there. It won't match Abnormal's behavioural focus on BEC, in my view, but it costs you nothing extra.
For a wider view of the market, our guide to AI cybersecurity companies covers the main vendors, and our roundup of cybersecurity AI automation tools looks at tools that cut analyst workload beyond email.
Verdict
I'd give Abnormal AI 4 out of 5. It's very good at the job it was built for, spotting the believable, link-free fraud emails that older gateways wave through, and it connects to Microsoft 365 or Google Workspace without rerouting your mail. Customer reviews back that up more consistently than I expected for a security product.
What holds it back is mostly money and maturity. The quote-only pricing, the flat platform fee and add-on modules make it hard to budget for, and half of the new AI Security suite was still in preview at launch. If you're a mid-size or larger company on cloud email and BEC keeps you up at night, I'd ask for the proof of value. If you're small, or still on-premise, skip it for now.
FAQ
What is Abnormal AI?
Abnormal AI is a cybersecurity company, formerly called Abnormal Security, founded in 2018. Its main product protects Microsoft 365 and Google Workspace email by learning normal behaviour for each person and vendor, then removing phishing, business email compromise and fraud messages that don't match. It now also sells identity and AI security products.
Is Abnormal AI free?
No. There's no free plan, no free trial and no public price list. Instead, Abnormal offers a proof of value that runs beside your current email security and shows which attacks it would've caught.
How much does Abnormal AI cost?
Pricing is quote-only. Published buyer deal data suggests roughly $20 to $35 per mailbox per year for the core email product, plus a flat platform fee of about $5,000 to $15,000, with add-ons on top. A 1,000-mailbox company might pay around $20,000 a year, but your quote could differ a lot.
Does Abnormal AI replace Microsoft Defender or a secure email gateway?
It can replace a separate gateway for many companies, since it works alongside Microsoft's or Google's built-in filtering via API. Some customers still keep a gateway such as Proofpoint in front of it. It doesn't replace Microsoft's own protection; it adds a behavioural layer on top.
Is Abnormal AI legit?
Yes. It's an established security vendor that says it protects more than 5,000 organisations, including 30% of the Fortune 500, and it's sold through AWS Marketplace. PeerSpot users rate it 9.4 out of 10 on average, though reviewers do mention false positives and high prices.
Ratings around the web
Ratings as of 20 Nov 2026. Scores change over time; follow the links for the latest.
What users are saying
We read about 22 recent business-user reviews: the G2 reviews mirrored on AWS Marketplace (September to November 2026), PeerSpot's review summaries and two Capterra reviews. Most praised fast, automatic phishing removal and easy setup; the mixed ones flagged misfiled legitimate mail, missed mail bombing, price and missing on-premise support. These are business software review sites, so the sample leans toward IT staff, not end users.
Common praise
- Fast, automatic removal of phishing and BEC emails
- Quick API setup with little configuration
- Cuts the time analysts spend triaging reported emails
Common complaints
- Some legitimate emails misfiled and occasional misses, such as mail bombing
- Premium, quote-only pricing that's hard on schools and smaller firms
- No on-premise email support, no outbound scanning and no mobile app
Based on: G2 reviews of Abnormal on AWS Marketplace, PeerSpot reviews of Abnormal Security, PeerSpot reviews of Abnormal Security (page 2), Capterra reviews of Abnormal AI
What we like
- Behavioural AI aimed at BEC, vendor fraud and link-free phishing that gateways miss
- Connects to Microsoft 365 and Google Workspace by API, with no MX record changes
- Automatic removal of malicious emails and calendar invites cuts manual triage
- Detection 360 turns reported misses into tracked, human-reviewed investigations
- Strong customer reviews, including a 9.4/10 average on PeerSpot
What could be better
- No public pricing, free plan or trial, plus a flat platform fee
- Some legitimate emails get misfiled and some attacks, like mail bombing, slip through
- No support for on-premise email and no mobile app
- Most of the new AI Security suite was still in private preview at launch



