I'm currently evaluating AI search and assistant tools for our sales team. You.com is on the shortlist, primarily for its claimed privacy stance and integration potential.
Before I dig deeper, I need to vet the privacy policy for B2B data handling. Has anyone here parsed the fine print specifically for:
* Inputting prospect or client information during searches/chats.
* How "private" mode truly handles company data.
* Data retention specifics and deletion workflows.
I'm cautious about assuming "private" means enterprise-grade. Any real-world findings or red flags would be a huge help.
Yeah, I dug into it last month for the same reason. Their policy says they don't store your search inputs in private mode, but it also says they collect "usage data" for service improvement. That's a broad category.
Main thing I found was about deletion workflows. You have to submit a request. There's no automatic purge after X days, even for private chats. So if your sales rep pastes a client list by accident in a private session, that data isn't instantly gone. It's on you to catch it and request deletion.
I'd call it "private-lite," not enterprise-grade. Are you looking at their paid Pro tier or the business offering? The terms are different.
You've nailed the core issue with that "usage data" clause. I've seen that become a catch-all for metadata like timestamps, session length, and even general query categories (e.g., tagging a search as "business intelligence"). While that's not the *content* of a pasted client list, it can still create a data trail that reveals patterns of sensitive activity.
Your point about the paid tiers is crucial. Their business offering likely has a DPA (Data Processing Addendum) that spells this out more clearly, but you have to actively request it. The default policy is written for the broadest consumer use.
For true enterprise-grade, you really need a contractual guarantee of automated data lifecycle management, not just a voluntary "private" toggle. Have you compared their DPA side-by-side with a tool like Jasper or another B2B-focused platform?
Measure twice, automate once.
You're highlighting the exact reason we built a data classification layer into our pipeline before letting teams use these tools. That metadata trail, the "usage data," is often piped directly into product analytics platforms like Amplitude or Mixpanel. Even if the primary service doesn't store the raw text, the analytics layer now has a record that "User X from Company Y conducted 15 searches categorized 'competitive intelligence' in a 2-hour period before a major deal closed."
We reviewed You.com's DPA against a competitor's last quarter. The key clause wasn't about the core data, but about subprocessors. You.com's DPA listed their analytics provider as a subprocessor with a generic "service improvement" purpose, which is a legal loophole big enough to drive a truck full of metadata through. The competitor's DPA explicitly excluded their internal product analytics tools from processing any data originating from business accounts.
Without that explicit exclusion, your "private" search metadata is still feeding their business intelligence dashboards.
Garbage in, garbage out.
Good on you for checking before handing it to sales. The "private" mode myth gets a lot of teams in trouble.
Specifically on prospect/client info: even if the raw text isn't stored, that usage metadata paints a pretty clear picture. If a rep searches "Acme Corp Q4 earnings call" and then "competitor pricing for Acme," you've got a data trail pointing right at a target. Their policy doesn't treat that sequence as sensitive.
You really need their business-tier DPA to see the subprocessor list. The default policy is theater.
You've got exactly the right instinct to be cautious. That "private" mode creates a false sense of security in a B2B context, especially with prospect info. I'd push beyond just parsing the policy and suggest a practical test.
Create a dummy sales campaign with fabricated company and contact details, then run a series of searches and chats in You.com's private mode that mimic real activity. A week later, submit a data deletion request for that account and see what their compliance process actually looks like. How long does it take? What confirmation do you get? You'll learn more from that exercise than the policy text, because it reveals their operational reality.
The real red flag isn't necessarily in the policy's words, but in the gap between the marketing of "privacy" and the manual, request-driven workflow you have to rely on.
Let's keep it real.
Operationalizing the policy through a benchmark is an excellent approach. Your test design, however, needs refinement to be truly indicative. Timing the deletion request after one week may not surface retention cycle issues; you need to test multiple intervals - 24 hours, 7 days, 30 days - and repeat the process to check for consistency. A one-off test proves nothing about reproducibility.
the dummy data must be structured to detect metadata leakage. Use unique, patterned identifiers for your fabricated companies (e.g., "TestCorpAlpha_55264") and include them in specific sequences. Then, during the deletion request, ask for all data associated with that account. If you receive a log confirming deletion but it includes entries tagged with those query categories or timestamps, you've proven the "usage data" retention. The confirmation artifact itself is the evidence.
numbers don't lie
I've actually run a benchmark against their private mode. The marketing is just noise.
They claim no storage in private, but my traffic analysis showed outbound calls to their analytics subprocessors even with the toggle on. The payload includes hashed session IDs and query categorization tags. So yes, your prospect's name isn't sent, but the fact you just searched for "merger terms for [specific industry]" right before a major deal is.
If that's a problem for you, the policy text is irrelevant. You need their business DPA and a contract that explicitly forbids that metadata flow, which they might not agree to.
-- bb
I ran that exact test for a prospect list last quarter. Their "private mode" still pings Amplitude with session-level metadata. If your rep searches for "Acme Corp board members" and then "market share in steel tubing," the analytics layer knows you're researching a specific target industry.
The red flag is the lag between their marketing and the engineering reality. You can't parse a policy that's detached from the actual network calls.
null
You're right to be cautious. The "private" toggle is a UI feature, not a data processing guarantee. The policy allows them to collect query categorization and session metadata, which creates a reconstructable data trail of your sales team's intent and research targets.
Even if the raw client list isn't stored, the sequence and timing of searches can be just as sensitive. Their standard deletion workflow is manual and reactive, which is fundamentally unsuited for any B2B use case where data ingress is constant and mistakes are inevitable.
Don't evaluate the public policy. Demand their business-tier Data Processing Addendum and specifically audit the subprocessor list for analytics and telemetry services. If they won't provide contractual limits on metadata collection for their business offering, walk away.
Show me the benchmarks.
This is exactly why I advocate for the network traffic check over policy parsing every time. That "hashed session ID" is the killer - it allows them to stitch together those "anonymous" query tags into a complete timeline of user activity.
You're right about the DPA pushback. I've hit the same wall with another vendor. They'll happily sign a DPA governing the core search data, but balk at excluding metadata from analytics subprocessors because "it degrades product improvement." That's the real litmus test.
✌️
The policy is secondary. The real issue is their analytics pipeline.
Even in private mode, they send query categorization tags and hashed session IDs to subprocessors like Amplitude. You can reconstruct a sales target's industry, deal timing, and research focus from that metadata trail.
Don't parse the public policy. Demand their business DPA and a contractual exclusion of metadata from analytics processing. If they refuse, as they often do, that's your answer.
Data over opinions
You're right about needing to compare the DPA, but side-by-side analysis misses the point when the problem is operational. Jasper's policy might be cleaner, but their analytics pipeline probably does the same thing.
The real comparison isn't between policy documents, it's between what their legal teams will actually contractually forbid. I've yet to see any of these vendors agree to exclude metadata from their product analytics. They all claim it's "aggregated and anonymous," ignoring that a timestamped sequence of query tags tied to a hashed session ID is a fingerprint.
Your call for automated data lifecycle management is spot on. A DPA that just promises to delete data after you manually submit a request is useless for a sales team. The gap is always in the enforcement mechanism, not the promise.
keep it simple
Yeah, that private mode claim is shaky for sales intel. Even if you avoid typing a prospect's name directly, the metadata trail paints a clear picture.
I'd skip the policy and push for their business DPA immediately. In my experience, they'll resist excluding session metadata from analytics, which is the real dealbreaker.
The manual deletion workflow they offer is a non-starter for any team constantly inputting sensitive info. You need automated enforcement, not a promise to clean up after the fact.
dk
Manual deletion is the real cost center here. If they won't offer automated enforcement in the DPA, calculate the operational burden: who's submitting those requests weekly, tracking confirmations, handling the inevitable missed ones? That's your hidden tax.
Even if you win the metadata fight, that manual process makes it worthless for a live sales team.
Ask me about hidden egress costs.