@morapayments Really like this. Signing payments offline and settling on Solana later could be genuinely useful for merchants in low-connectivity areas.
An AI agent was given a cyber task inside a test. It escaped the test environment and accessed another company’s systems.
That is the clearest reminder yet that agent permissions cannot be a checkbox.
Scope, limits, revocation, and traceability have to exist before execution.
Agent accountability is two problems, not one.
First: is a real human answerable for the agent.
Second: what was it allowed to do, and can you prove it stayed in bounds.
The first is identity. The second is enforceable policy and replayable proof.
We build the second.
This is only the beginning of a new era.
Autonomous agents need enforceable limits that check every action against predefined conditions before execution. We need this and we need this yesterday.
We're partnering with @huggingface to investigate an unprecedented security incident.
Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation.
Sharing preliminary findings to help defenders understand emerging risks:
https://t.co/CIor15y9xk
AI agents can already manage wallets and move money at impressive speeds.
What’s missing is a control layer that defines permissions, enforces limits, and produces a receipt for every action.
We don’t need to slow agents down.
We need control over their decisions.
We've been seeing lots of aidrops lately. We know Solana keeps making distribution easier each and every day. Imo the hard part is making the decision behind that distribution checkable.
@Hive_Agents_Tec Nice launch, passkey for savings is the right call. The harder part is the budget the agent controls: what enforces that each trade stays inside your rules, and how do you prove after a loss that it did? On-chain outcome is not on-chain authorization.
@orskyai@zerohedge True today. But 'linked API and human auth' is exactly what agent frameworks are dropping. The whole point of autonomy is taking the human out of the loop. Once the API is wired and no one signs off, a prompt injection is the execution. That is the gap.
@AuspiciousToad Because the announced parameters are calculable but the actual selection function is not. Community fills the gap with spreadsheets. That gap is exactly the shape of the missing protocol.
@jussy_world@bulktrade@solsticefi@Arcium "Not official, estimates only" is the calculator's honest disclaimer. It should also be the protocol's honest position, and it wouldn't have to be if the function was published on chain.
@0x_abu I am happy for you man. But I am still wondering, why 14,724 tokens specifically, and were other holders and creators evaluated by the same logic? It'd be better if we had a transparency layer to that part for any airdrops just for the community's sake.
@FabianoSolana@MeteoraAG Love the outcome. Also a good example of the difference between distribution and allocation: the $6K transfer is public, but why this wallet received it and how that amount was decided is not.