Your crypto inventory is finished.
Your agents just made it wrong.
Post-quantum migration is an inventory problem before it is an algorithm problem — most organisations will spend 2026 and 2027 mapping what cryptography they have. A server estate holds still while you map it. An agent fleet does not: it creates keys, sessions and channels continuously, so the inventory is obsolete the moment it is finished. PQ CENSUS measures the fleet again, every run, from evidence the platform already holds — and prints what it could not measure next to what it could. The census is taken for the superintelligent AI agents that will sign, pay and attest with these keys — agents whose action we would not accept on their word.
Four measurements per agent — three states each, never a silent pass
Is the agent’s harness configuration post-quantum signed?
Every configuration we govern carries an ML-DSA-87 (FIPS 204, level 5) signature, verified on each run.
What does the agent authenticate with?
Every credential the agent uses is measured against the post-quantum bar. An agent with nothing registered is reported as unmeasured, not clean.
Can the destinations it reaches do TLS 1.3 at all?
A destination that cannot carry post-quantum key exchange is reported as such, whatever the agent’s keys look like.
Does the destination actually negotiate X25519MLKEM768?
Observed on the wire, never inferred from a vendor’s roadmap.
The house rule: when a higher level exists, we take it. Our signatures are ML-DSA-87 — FIPS 204 level 5, the highest — and our symmetric bar is 256 bits, above the NIST minimum we quote. READY means the house grade on every dimension, not the minimum on any.
Four measurements per agent — signature, credentials, channel, hybrid key exchange — each observed, never inferred. Unmeasured is reported as unmeasured. We ran it on our own house before we offered it to anyone: some of the largest edges on the internet already negotiate the hybrid today, and some widely used APIs do not yet. The census says which, per agent, per destination — and it says UNVERIFIED where it could not look.
The dates — quoted from the primaries
We quote the minimum. We use the highest.
FIPS 140-2 goes historical
“All FIPS 140-2 certificates are placed on the Historical List.” From here on, a module validation means FIPS 140-3. NIST CMVP transition schedule.
Industry pulls the date forward
A major internet infrastructure provider now targets 2029 “to be fully post-quantum (PQ) secure including, crucially, post-quantum authentication.” Its roadmap, 7 April 2026.
The US executive order
High-value and high-impact federal systems move to PQC for key establishment by 31 December 2030 and for digital signatures by 31 December 2031. Executive order, 22 June 2026.
An agent authenticates many times a day; its identity is machine identity, and machine identity is moving to post-quantum signatures by mandate. PQ CENSUS says, per agent, what it signs with today.
Measured, not configured: the group your agent negotiated, the library it linked, the scheme it signed with, and whether a regulator still recommends it.
The credential the agent presents is graded too. Any agent credential, whoever issues it, is graded by the scheme that signs it. Of the two agent-passport launches of 15 October 2026, neither named its signature scheme and neither mentioned the word quantum. If your agents carry a passport from a platform, we will tell you what it is signed with.
Algorithm diversity is the policy
An AI broke HAWK in about 60 hours
An AI lab’s model found a new attack on the HAWK post-quantum signature scheme in “about 60 hours in total”: the expected cost of full key recovery against HAWK-256 “was thought to be 2^64 but was demonstrated … to be 2^38”. HAWK was withdrawn from NIST’s additional-signature process. The research · NIST round 3.
ML-DSA-87 and ML-KEM untouched
The researchers state the attack is specific to HAWK and “does not impact other NIST post-quantum signature candidates or lattice-based cryptography in general.” Our house grade stays ML-DSA-87 — and we never rest a fleet on one scheme. A change of algorithm is a re-sign, not a re-architecture.
The playbook: when a scheme is withdrawn, withdraw it for new signatures, keep the hybrid for what already exists, and map the fleet in an hour, because the census already knows what each agent signs with. Our own receipt is signed with ML-DSA-87 today, so it needs no migration.
What you get
Fleet readiness, with its denominator
How many agents are READY, PARTIAL, CLASSICAL, UNVERIFIED — and how many the run actually measured. Re-run daily; the fleet moved since yesterday.
Harvest-now, decrypt-later, dated
Which agents move data that is still sensitive in 2035 over anything not proven post-quantum. The CFO-legible version of the threat: a function of data lifetime, not of when the machine arrives.
Signed with ML-DSA-87 — it outlives the audit
Every run is hashed and post-quantum signed. Crypto-agility is attested, not promised: a change of algorithm is a re-sign, not a re-architecture.
Why nobody else sells this
Because nobody else holds the agent census to build it on. A crypto-inventory vendor sees certificates and keys on servers. AIM sees the agents — which one holds which credential, which destinations it was seen reaching, whether its harness is signed — and can therefore say, per agent, what is post-quantum and what is not, today, and again tomorrow.
PQ CENSUS is part of AIM — €1,350 / €6,750 / €32,400 per agent per year by passport class, minimum €135,000
One conversation with whoever owns your post-quantum migration plan. We bring the honest number and your fleet’s first census.
Talk to us AIM pricing · the passport ladder