Most people in crypto believe the greatest threats are smart contract bugs, private key leaks, or exchange hacks. They are wrong. In 2026, the most efficient attack surface is the corporate communication layer. On August 31, Palo Alto Networks Unit42 exposed a campaign named Spring Ring that used Microsoft Teams' default external access to compromise over 150 employees across ten organizations. The attackers didn't need a zero-day. They needed a registered .onmicrosoft.com tenant and a convincing voice. For crypto firms, this is not a distant corporate IT problem. It is a direct threat to the people who hold the keys.
The campaign ran from January to April 2026, attributed to Cloaked Ursa, a group linked to APT29 and Muddled Libra. The attack chain is deceptively simple: a Teams chat, a vishing call, a remote management tool, an NTLM relay, and finally domain-wide privilege. Unit42's analysis identifies two distinct campaigns. Campaign A deployed RMM tools and an obfuscated PowerShell RAT, disabled AMSI, and performed local enumeration. Campaign B was more targeted: S3-hosted payloads named for victim companies, persistence mechanisms, headless Edge browser hijacking, and PetitPotam NTLM relay to hit domain controllers. The endgame: Pass the Hash (T1550.002), domain admin.
This is an identity-layer attack, not a network-layer one. The initial entry is social engineering through a trusted collaboration platform. Microsoft Teams external access is enabled by default in most M365 tenants. That means any external user with a free tenant can send a chat to an internal employee. In this campaign, attackers impersonated IT help desks with display names like ITProtectionDepartment. The trust is weaponized. The vulnerability is not in the software; it is in the configuration and the human psyche.
I learned a similar lesson in 2017 while auditing a token distribution contract. The critical bug wasn't in the token logic; it was in an integer overflow in the distribution script. The default conditions were the flaw. Teams external access is the distribution script of enterprise identity. And just as I urged the team to validate all arithmetic before launch, I now urge every crypto institution to audit their collaboration defaults. The default configuration is the vulnerability, not the code.
The macro context amplifies the risk. In my work modeling Bitcoin ETF inflows, I saw that institutional adoption brings a new layer of fragility: the operators. The same centralized infrastructure that handles fiat now handles crypto custody, communication, and key management. Spring Ring demonstrates that the chain is only as strong as the weakest hand. Unit42 Cortex telemetry shows that collaboration tool phishing now represents 42% of all phishing alerts, up from 30% in the previous four months. KnowBe4 reports a 41% increase in Teams-based attacks. Exchange Auth Bypass CVE-2026-62911 leaves 21,899 unpatched servers exposed. The risk landscape is expanding, and crypto firms are not isolated.
The technical path of Spring Ring is worth dissecting because it reveals the brittleness of hybrid identity architectures. The chain: Teams chat → vishing → RMM access → NTLM relay → Pass the Hash → domain-level permissions. The players trusted the caller because the Teams channel is considered internal, even though it came from an external tenant. Then they granted remote access to a fake help desk. The attacker then moved laterally through Windows domains, abusing NTLM relay and PetitPotam. PetitPotam has been patched since August 2021. Yet here it is in a 2026 campaign, because organizations fail to disable legacy authentication protocols. The attackers are not seeking novel exploits; they are harvesting known weaknesses in enterprise identity.
For crypto, the lesson is twofold. First, private keys live on workstations that are joined to these domains. If an attacker achieves domain admin, they can reset passwords, install keyloggers, or directly access memory of signing processes. The blockchain is immutable, but the cold wallet is only as cold as the air gap. Second, the social engineering aspect is far more dangerous than any phishing email. Voice calls on Teams carry an implicit trust that email has lost after years of training. The human element is the new smart contract: it can be exploited for function calls with unexpected side effects.
Let me give you an example from my own experience. In 2020, I built a Python risk model to evaluate Uniswap V2 liquidity pools. I allocated $500,000 into Aave and Compound with hedging. My report predicted stablecoin depegging due to collateral opacity. I exited two weeks before bUSD collapsed. The lesson was that macro fragility often hides in seemingly stable settings. The same logic applies here: the Teams default setting is the stablecoin of collaboration. It holds until someone discovers the arbitrage.
Now the contrarian angle. The crypto maximalist narrative claims that decentralized systems are secure because they are trustless. That is a fallacy. The infrastructure surrounding crypto is deeply trust-based: employees, tools, and communication platforms. The decoupling thesis—that crypto markets move independently of traditional finance—does not apply to operational security. If a domain controller is compromised, the attacker can reset passwords, access email, and potentially steal the private keys stored on joined workstations. The blockchain might be permissionless, but the employees are not. The incentive structure of attackers is clear: they go where trust is highest and defenses are weakest. Incentives break before code does.
Some might argue that Teams attacks only affect traditional enterprises, not crypto-native firms. But crypto firms are prime targets. Nation-state actors like APT29 are after strategic intelligence, yet they are also drawn to digital assets. The S3-hosted payloads with victim company names show a level of preparation that speaks to targeted surveillance. If you have employees with access to hot wallets or governance keys, you are on the radar. And consider this: on-chain governance voter turnout is perpetually below 5%. The same apathy applies to security configurations. The majority of Teams users have never reviewed their external access settings. The principal-agent problem is real: security is not the user's incentive, so it gets ignored.
The 2022 Terra-Luna collapse taught me that mechanism design often fails under stress. Terra's anchor protocol promised 20% yields, and the mathematical inevitability of the death spiral was visible. Similarly, Teams default external access promises frictionless collaboration, and the human trust curve is just as predictable. The market is sideways right now, and firms are focused on positioning, not security. That is a mistake. Chop is for positioning, but also for preparation. The quiet period is the best time to harden identity infrastructure.
So what should crypto firms do? First, restrict Teams external access to a whitelist of trusted domains. If you don't use cross-tenant collaboration, disable it. Second, define a clear protocol for IT help desk interactions: no unsolicited remote access, no voice calls from unknown numbers. Third, disable NTLM everywhere possible, or at least deploy the security zones that block relay attacks. Fourth, implement phishing-resistant MFA (FIDO2/WebAuthn) for all privileged users. Fifth, deploy identity threat detection and response (ITDR) that monitors for pass-the-hash and abnormal NTLM activity. These are not complex solutions, but they require the same rigor you apply to smart contract audits.
Microsoft, on its part, must re-evaluate its default stance. The company describes this attack pattern in a security blog from April 2026, but the default remains open. Just as we demand that DeFi protocols change default parameters to protect users, we should demand that enterprise platforms do the same. The platform is the new financial infrastructure. Third-party security firms like Palo Alto are stepping in, but the first line of defense is the platform itself.
The takeaway is not to abandon collaboration tools; it's to recognize that every layer of abstraction introduces trust assumptions. The blockchain removes the need to trust the counterparty, but it does not remove the need to trust the operating system, the domain administrator, or the human who answers a phone call. Volatility is the tax on uncertainty, but the real tax is the human layer. In this cycle, the most effective hedge is not an options strategy—it's operational hygiene. The attacker works the path of least resistance. Don't make your Teams chat that path.
I have seen enough cycles to know that after every major compromise, the industry says they should have known. In 2017, the lesson was to audit smart contracts. In 2020, it was to stress-test yield. In 2022, it was to question algorithmic pegs. In 2026, let it be this: the attack surface extends to the collaboration layer. The private keys are only as secure as the conversation you had this morning.