A Cautionary Tale from 2025/2026
Here’s something I never thought I’d have to write about, but after watching it happen three times in a single year (really more than that), it’s time we talked about it openly.
I’ve now seen three separate VC seed deals in 2025 where the founders simply decided to keep the money for themselves and not build the product. Not fail trying. Not pivot. Not even give it a real shot. Just… took the money and walked away.
All three were structured on SAFEs. And in each case, the founders took the position that they owed investors nothing. That the SAFE holders weren’t even really “investors” at all. Just… SAFE holders with no rights, no recourse, no standing.
Let me be clear: I didn’t personally invest in these deals, though I easily could have. They all had a lot going for them—or so it seemed at the time. Strong team credentials. Compelling market opportunity. The kind of deals that get circulated in good investor networks.
And here’s what makes this even more sobering: These weren’t small checks. We’re talking $2M-$5M+ in seed SAFEs. This isn’t angel round territory. This is serious institutional money.
The Playbook (If You Can Call It That)
Here’s how these situations unfolded:
Deal #1: The $5M+ SAFE with No Reference Checks
This was a $5M+ seed round on SAFEs. Multiple investors. And here’s the kicker: No reference checks were done with the founders’ prior investors. At least not on all the founders.
In the hot deal environment, investors moved fast. FOMO was real. The round came together quickly. Everyone assumed someone else had done the diligence.
Spoiler alert: No one had.
Deal #2: The $4M Split Scenario
This one raised around $4M in seed SAFEs. Reference checks were done… but only on one founder. The founder who didn’t keep any of the money.
The founder who wanted to keep it all? No reference checks done.
Think about that for a second. $4M deployed, and the person who ends up being the problem wasn’t properly vetted.
Deal #3: The $2M+ No-Lead Round
This company raised $2M+ but the checks were all small with no lead investor. And as far as I can tell, no one did reference checks at all.
Classic party round problem: everyone’s check is small enough that deep diligence doesn’t feel worth it. Everyone assumes someone else is doing the work. No one actually does.
The outcome? One founder eventually returned some cash after keeping a large chunk for themselves. Called it “expenses” and “market research.” The investors who pushed back got maybe 40-60 cents on the dollar. Those who didn’t push got nothing.
Another hired an attorney who told them plainly that nothing was owed to the investors and that the founders could pay themselves whatever they wanted out of the SAFE proceeds. They’re keeping it all. Full stop.
The third is still unfolding, but following a similar pattern.
None of these involved fraud in the traditional sense. No fake revenue. No fictional customers. Just founders who raised on SAFEs, looked at the risk/effort/reward calculus, and decided the best ROI was doing nothing and keeping the cash.
Why This Is Happening Now
This is the flip side of everything exploding so quickly in the Age of AI.
When you have deals going from $0 to $10M ARR in 12 months, when you have companies getting to unicorn status in 18 months, when the speed of everything has accelerated by 10x—you also get the inverse scenarios playing out at similar velocity.
The same market dynamics that enable hyper-growth also enable hyper-opportunism:
- Capital is abundant for AI deals, making it easier to raise without deep diligence
- FOMO is real among investors, creating pressure to move fast on deals
- Technical barriers are lower with AI tools, so it’s easier to look credible without much substance
- The complexity of AI makes it harder to assess what’s real vs. vaporware in early stages
- Hot rounds close fast, and diligence gets skipped in the rush
Add to this the reality that many investors—even sophisticated ones—don’t fully understand the legal mechanics of SAFEs. They think of themselves as “investors” because they wired money. But legally? That’s not always what the documents say.
The Reference Check Problem
Let’s talk about what went wrong with diligence in these deals, because the thread responses raised an important question: Were reference checks done?
The answer varies, and it’s instructive:
Case 1 ($5M+ round): No reference checks done with founders’ prior investors. Hot deal, moved fast, investors assumed others had done it.
Case 2 ($4M round): Reference checks done on one founder (who didn’t keep the money) but not on the other founder (who wanted to keep it all). Classic mistake: checking the wrong person.
Case 3 ($2M+ round): Party round, no lead, checks were small. As far as I can tell, no one did reference checks.
Here’s the hard truth: Reference checks are important, but they aren’t perfect.
When deals are hot and happen fast, many investors skip them—at least for the most part. They tell themselves:
- “It’s just a $100K check, not worth the time”
- “The lead investor probably did them” (when there is no real lead)
- “This round is closing Friday, no time”
- “The other investors are smart, they must have done diligence”
And when everyone thinks someone else did the work, no one actually does.
But even when you do reference checks, there are failure modes:
- You check the wrong co-founder. Like in Case 2, where they vetted the ethical one.
- Prior investors give positive references because they don’t want to hurt their own portfolio company. If a founder is still running a company they invested in, they’re not going to badmouth them.
- The founder hasn’t shown their true colors yet. Maybe this is their first time pulling something like this. Past performance isn’t always predictive of future bad behavior.
- Reference checks focus on competence, not character. “Are they a good engineer?” is a different question than “Will they steal the money?”
Still, skipping reference checks on a multi-million dollar investment is inexcusable. Especially in 2025, when these patterns are now known.
The Math That Enables This
Let’s run the numbers on what a bad actor founder can do with a $2-5M seed round on SAFEs:
Example: $3M Raise
- Pay yourself $200K/year for 2 years: $400K
- Hire 2-3 contractors for 6-12 months: $300K
- “Product development” (minimal AWS, some SaaS tools): $100K
- “Marketing expenses” (a few conferences, some ads): $50K
- “Customer development” (impossible to audit): $100K
- Legal and incorporation costs: $50K
Total “spent”: $1M
Remaining to keep: $2M
You “tried” for two years. The market wasn’t ready. The timing wasn’t right. The technology wasn’t there yet. Customer feedback indicated a pivot was needed but the team didn’t have the right DNA for that new direction.
All true statements that are impossible to disprove.
And you walk away with $2M+ for two years of “work” that may have involved very little actual work.
That’s better than a Big Tech senior engineering salary, with way more upside optionality and way less accountability.
The SAFE Problem Nobody Talks About. But SAFE’s Aren’t The Issue, Per Se. Just An Enabler.
SAFEs were designed to be simple. And they are—simple for founders. But that simplicity comes with trade-offs that most investors don’t think about until it’s too late.
Here’s what a SAFE actually is: It’s an agreement to receive equity in a future priced round. Until that priced round happens, you don’t own equity. You’re not a shareholder. You don’t have board rights. You don’t have information rights (unless explicitly added). You don’t have much of anything.
The legal term is “contractual right to future equity.” That’s very different from being a shareholder.
So when founders decide not to do a priced round? When they decide not to build the company? Your SAFE doesn’t convert. And depending on how it was structured, you may have zero legal recourse.
The attorneys in Deal #2 weren’t wrong. Under many SAFE structures, founders legally can pay themselves whatever they want from SAFE proceeds, as long as they’re not committing fraud. If the SAFE doesn’t specify limits on founder compensation, or use of proceeds, or information rights, or dissolution preferences—and many don’t—then there’s no contractual violation.
You just have a very expensive lesson in contract law.
What Changed?
I’ve been in the venture business for 12+ years, managing $200M across three funds. I’ve seen SAFEs since their inception. They aren’t the issue per se. This is new.
Previously, the failure mode was: founders tried hard, market didn’t work, company shut down, everyone got nothing. That’s venture capital. That’s the risk we all sign up for.
But the social contract was intact. Founders were trying. They were putting in the work. They were making a genuine effort to build something.
What’s different now is the explicit calculation some founders are making: “I can raise $3M on SAFEs for Hot AI Startup, pay themselves a $200K salary for two years while ‘exploring’ the market, and if nothing clicks, I just keep what’s left. No accountability. No board. No problem.”
That’s not failing. That’s a different business model entirely.
Why This Matters More in AI
AI deals are particularly vulnerable to this dynamic because:
- It’s genuinely hard to tell what’s real in the early stages. Is this proprietary technology or just a wrapper around Claude/GPT-4?
- The skill gap is real. Many investors can’t personally assess the technical depth of an AI product, making it easier to fake momentum.
- Customer development takes time. It’s plausible that founders spent 12-18 months “talking to customers” and “iterating on the product” without having much to show for it.
- Demos are cheap. You can build impressive-looking demos in days now using tools like Replit. But demos ≠ product ≠ business.
- The pivot excuse is always available. “We started in vertical X but learned the real opportunity is in vertical Y” can cover a multitude of sins.
- AI infrastructure costs are real but opaque. It’s easy to claim you spent $500K on compute and model training. Who’s going to audit your AWS bill in detail?
What Investors Should Do
If you’re writing checks into seed deals, especially on SAFEs, here’s what’s now required diligence—even for small checks in party rounds:
1. Always Do Reference Checks. Always.
Not just on the CEO. On every founding team member who has significant equity and spending authority.
And don’t just check with their prior investors (who have conflicts). Check with:
- Former colleagues who don’t have money at risk
- People who worked for them, not just with them
- Customers or partners from previous ventures
- People who can speak to character, not just competence
Budget 3-5 hours minimum for thorough reference checks on a $2-5M seed round. If you don’t have time for this, don’t write the check.
2. Never Assume Someone Else Did the Work
This is especially true in party rounds with no lead investor. If every investor is writing a $50-250K check, there’s a high probability no one is doing deep diligence.
Either:
- Organize with other investors to split diligence workstreams
- Hire a third party to do background checks
- Or accept you’re making a bet with materially higher risk
3. Add Protective Terms to Your SAFEs
Don’t use the standard Y Combinator SAFE without modifications. Add:
- Information rights: Monthly updates minimum, financials quarterly
- Pro-rata rights: Ability to participate in future rounds
- Founder salary caps: Reasonable limits on cash compensation (e.g., $150-180K for seed stage)
- Use of proceeds restrictions: Broad strokes on what the money can’t be used for
- Dissolution preferences: SAFE holders get their money back before founders in a wind-down scenario
- Spending velocity limits: Can’t burn >$X per month without investor notification
Yes, this makes SAFEs more complex. That complexity is now necessary. The instrument has been exploited enough that the original simplicity is a bug, not a feature.
4. Verify Everything Technical
Don’t rely on the pitch deck. Verify:
- GitHub activity (is code actually being written? by whom?)
- Technical architecture (bring in someone who can assess if this is real or a thin wrapper)
- Model training logs (if they claim to have trained custom models, prove it)
- Customer conversations (talk to the customers they claim to have)
- Product depth (can you actually use the product, or is it just a demo?)
For AI deals especially, technical verification is non-negotiable at the $2M+ level.
5. Milestone-Based Funding
Instead of cutting a $3M check upfront, structure it as:
- $1.5M now to prove initial traction
- $1.5M at 9-12 months upon hitting agreed milestones (X customers, Y revenue, Z product capabilities)
This doesn’t eliminate risk, but it forces founders to actually do something to get the second tranche. And it dramatically reduces the amount of money available to simply walk away with.
6. Syndicate with Operational Investors
Make sure at least one investor in the round is:
- Technical enough to assess product progress
- Available enough to engage monthly with founders
- Experienced enough to know what real vs. fake progress looks like
- Willing to be the “bad cop” if things go sideways
The best defense against this behavior is active, engaged investors who founders know are watching.
The Broader Implications
One of the responses in my Twitter thread put it well: “When deals are hot and happen fast, many skip them, at least for the most part.”
That’s the meta-problem. The speed at which capital is moving in AI has created systematic gaps in diligence. And bad actors are finding those gaps.
Some other smart takes from the thread:
On trust and systems: “Trust without verification isn’t trust. It’s just hope. VCs are trading really fast for permanently destroying their reputation and future earnings potential. I had a founder reach out for an angel check like this to cover living expenses, and I roped them out of it.”
The reality is that the venture ecosystem runs on repeated games and long-term reputation. But that only works if:
- Information flows (investors talk to each other about bad actors)
- Consequences exist (bad founders get blacklisted)
- The timeline is long enough that reputation matters more than a quick cash grab
In the Age of AI, with capital moving at warp speed and new investors entering the ecosystem daily, those conditions don’t always hold.
On structural incentives: “People cash in quick buck for permanently destroying their reputation and future earnings potential. However, not everyone has the VC world’s ruthless portable companies, but, haven’t we learned at least from lesson VC world years about best 3 practices, it’s not.”
The question is: do founders believe the “permanent reputation destruction” threat anymore? If you can walk away with $2M after two years, and you’re not planning to be in the VC-backed startup world again anyway, maybe the trade-off is worth it to them.
That’s a depressing calculus, but we have to acknowledge it’s real.
The Path Forward
This isn’t a reason to stop investing in AI. The opportunities are real. We’re seeing portfolio companies grow faster than anything in the 2010s.
But it is a reason to level up diligence and tighten up documentation.
The speed and abundance of capital in AI has created new failure modes we haven’t seen before. Being aware of them—and protecting against them—is now part of the cost of doing business in this market.
Three deals totaling $10M+ just in my ecosystem in one year is a pattern. It’s not an anomaly anymore. It’s a risk factor that needs to be priced in and protected against.
Have you seen similar situations in your portfolio or deal flow? What protective measures are you putting in place for 2025 seed investments?
