Shai Kritz

models are unaligned, in the sense that they consistently exhibit internal biases toward themselves and also reason and justify crossings of moral boundaries in service of rewards or objective completion. historically, cyber posture started as static policies. allow / block lists, IP filtering, broad-reaching network policies, and sentinels at the door of data stores, DBs, or sensitive credentials. the slope of general, sophisticated offensive cyber capability is extremely steep. in reviewing these incident reports the most impressive (scary?) thing that stands out is the ability of agents/models to daisychain what otherwise seem like sensitive low-medium security findings into criticals so quickly. a) very obvious that human operators today cannot keep up b) static policies and heuristics are a speed bump, not a roadblock because of this, i think agentic defenders and net new models will be very big. we'll need solutions that are credibly neutral, uninvested in the ongoing agent work stream, and that don't care about the rewards at hand. so effectively, fighting ai with ai. defender agents will be token hungry, bc they'll need to effectively run inference on inference, if they're to oversee all agents running within a fleet. this feels a bit dystopian, very much like a pharmaceutical company that creates the disease and sells the "cure", but i don't really see another path forward. how else do you defend yourself against probabilistic geniuses in high-stakes environments with no structural guarantees? constitutions and best-effort promises of alignemnt are not enough.

Tokenmaxxing is easy when you're getting started. Modeled @chaoslabs AI usage at scale and saw a projected $150M/yr in spend. While I knew what it would cost, I had no real visibility into its impact, efficiency, and value. Sharing our journey on measuring AI spend ROI.

Over the weekend, we identified an attack on Chaos Labs. The surface area was strictly contained to operational wallets we use for routine onchain operations. At no point was the Chaos Oracle Network breached or compromised. Chaos Oracles run in a fully isolated environment with nodes distributed globally, protected by layered security and cryptographic controls. The oracles continued to publish prices across every network throughout the incident. We allocate a substantial share of our operating budget to cyber defense, alerting, and detection. These detection systems alerted us within seconds of suspicious activity, and we immediately moved to full lockdown. Our incident response policy treats any threat as the worst-case scenario by default, given the value secured by the Chaos infrastructure. With the backdrop of recent exploits, we triggered our highest-severity incident response from the moment we were alerted and notified relevant partners. The authorities and cyber professionals working with us have characterized the activity as consistent with nation-state attacks. As a precautionary measure, we rotated all keys, and have not observed any suspicious activity since. The investigation continues, and we will share more as it allows.

$200M was drained from Aave's rsETH markets following an exploit of Kelp's LayerZero bridge earlier today. The exploiter deposited rsETH, and borrowed 83,427 WETH and wstETH from Aave. The breakdown of extracted assets across Aave instances is: Ethereum 52,834 WETH Arbitrum 29,782 WETH 821 wstETH At the time of writing: - Aave rsETH markets are frozen. - No other LayerZero OFT token was affected. We are assisting multiple teams in investigating the root cause and the full path the funds took after extraction. A longer investigative report will follow soon.

Great conversation with @DefiantNews on Resolv + contagion. When yield optimization takes priority over risk management, exploits become a question of when, not if. Real-time risk oracles, monitoring, and alerts all exist. The industry just needs to start demanding it.

Chaos AI analyzed the vaults impacted by the USR exploit. A ton of interesting weekend transactions. Here, we build a knowledge graph, filtering for the Gauntlet USD Alpha Vault and the Gauntlet Resolv USDC vault: Pre-exploit: • USD Alpha had allocated ~438,440 USDC to the Gauntlet Resolv USDC vault • That vault was deposited into the impacted Morpho Resolv markets via the Gauntlet Resolv USDC vault. app.morpho.org/ethereum/vaul… • USD Alpha held receipt tokens representing this exposure Post-exploit, March 23: • 00:30 UTC, Gauntlet USD Alpha sends ~438,440 USDC worth of resolvUSDC receipt tokens (~405,439 shares) to a Safe. • 00:35 UTC, Gauntlet USD Alpha receives USDC ($438,401) from Coinbase. Receipt tokens out, USDC in, 5 min apart, w/ roughly the same notional. etherscan.io/tx/0x7db422e956… etherscan.io/tx/0x35ce2a750e… A ton more to dig into. Analyzing and contrasting vault curator operations and allocation patterns in real time is one of the use cases we’re building Chaos AI for.

1/ stETH CAPO Misconfiguration Today, a misconfiguration on Aave's CAPO oracle caused wstETH E-Mode liquidations, resulting in a loss of 345 ETH. No bad debt was incurred, and all affected users will be fully reimbursed. More below.

Chaos AI zero-shotting a 19-step sequence to open a 6x levered stETH/ETH loop on @aave V3 E-mode. In a sea of crypto x AI hype, the most differentiated and valuable use case is building agents that can act fully autonomously. A wallet, capital, and permissionless access to any onchain financial instrument/asset class are unique to our industry. In closed beta. DM me or sign up for early access.