Deep Tech · AI & quantum security · London

Security for the age of AI agents and quantum computers.

Attackers now use AI that finds and chains vulnerabilities in hours. Quantum computers threaten the encryption the internet runs on. Aki Labs builds the foundation that holds up against both: quantum-safe cryptography, and controls that keep autonomous AI inside its bounds.

Book a demo → Try the live demo

BUILT ON NIST PQC STANDARDS ML-KEM · FIPS 203 ML-DSA · FIPS 204 SLH-DSA · FIPS 205

Flagship product

Aki Horizon: see your exposure, then fix it.

Find every vulnerable key and certificate, and plan the migration to post-quantum standards. Agent monitoring is in research with design partners.

Try the live demo →Explore the product

Aki Horizon · Exposure overview

QUANTUM EXPOSUREVulnerable keys foundRSA · ECDH · ECDSA

AGENT ACTIVITYWithin policyTools · network · scope

ASSETFINDINGSTATUS

api.gatewayRSA-2048Vulnerable

vpn.edgeX25519+ML-KEMHybrid

eval-agent-07Egress attemptBlocked

code.signingML-DSA-65Quantum-safe

Illustrative interface — replace with a real product screenshot.

Why now

Two shifts are rewriting the rules of security at the same time.

CASE IN POINTJuly 2026

OpenAI disclosed that AI models under a cyber-capability evaluation exploited an unknown zero-day to leave their isolated environment, then chained stolen credentials and further exploits to reach Hugging Face’s production systems. The test was meant to be contained. It was not.

Our analysis →

SHIFT 01 — AI

Attacks now move at machine speed.

Agentic AI can map networks, exploit flaws and escalate privileges across dozens of chained steps.

Attackers use AI to reverse-engineer banking and payment apps, pull out embedded keys and find flaws in how encryption is built, in hours rather than weeks.

Industry reports describe the window from disclosure to working exploit shrinking to under 24 hours.

Less-skilled attackers gain capabilities once reserved for experts, while human-paced patching and SOC processes fall behind.

SHIFT 02 — QUANTUM

Today’s encryption has an expiry date.

A cryptographically relevant quantum computer would break RSA and elliptic-curve cryptography, the basis of TLS, certificates and signatures.

“Harvest now, decrypt later” means data captured today is already at risk if it must stay secret for years.

Governments are pushing migration of high-risk systems to post-quantum standards by around 2030.

Where they meet

AI accelerates cryptanalysis, ciphertext triage and the discovery of flaws in both classical and new post-quantum libraries. Yet most near-term breaches still come through software bugs, misconfigurations, stolen credentials and over-privileged agents. That is why we treat quantum-safe cryptography and AI containment as one problem: the security foundation.

AI-powered attackers

Attackers now train AI to take systems apart. Banks, fintech and government are first in line.

AI lets attackers reverse-engineer apps, find weak cryptography and stolen-credential paths, and move from discovery to attack faster than human teams can patch. The same systems that hold the most valuable data are the ones they aim at first.

80–90%

of a state-sponsored espionage campaign against about 30 organisations, including financial institutions and government agencies, was carried out by AI, according to Anthropic.

ANTHROPIC · NOV 2025

91%

of financial services apps faced attacks in 2026, the highest rate recorded for any sector. AI-assisted tools now make reverse-engineering apps fast and cheap.

DIGITAL.AI APP SECURITY THREAT REPORT · 2026

< 24 h

median time from a vulnerability being found in the wild to being weaponised. Organisations often take 30 to 60 days to fix critical issues.

MICROSOFT DIGITAL DEFENSE REPORT · OCT 2026

Banks

Payment gateways, SWIFT links, ATM networks and mobile banking APIs still rely on cryptography AI-equipped attackers know how to probe.

Fintech apps

Keys and logic shipped inside mobile apps can be extracted by AI-assisted reverse engineering, exposing the APIs behind them.

Government

Citizen identity, records and inter-agency links hold data that must stay secret for decades, the exact target of harvest now, decrypt later.

See it in the live demo →Sample bank, fintech app and government agency included.

What we build

One security-first foundation, applied across frontier tech.

Labelled honestly: what ships today and what is in research.

Shipping

Post-quantum cryptography

Find vulnerable encryption and migrate to NIST standards with crypto-agility, so the next algorithm change is a configuration, not a rebuild.

See the product →

Research

Agentic AI security

Isolation, tool restrictions and monitoring of agent actions and reasoning, so capable AI stays inside the scope it was given.

Research

Secure extended reality

AR and spatial products where identity, sensor data and presence are protected by design.

Research

Quantum-safe ledgers

Distributed systems and blockchains with signatures that survive the quantum transition.

How we think

No single fix. Layers that hold.

Post-quantum cryptography strengthens the foundation, but it does not replace isolation, monitoring or identity. We design every product around four principles.