Insights / Quantum and AI, 2027
Quantum and AI, 2027

Human speed in 2027: quantum deadlines meet the AI exploit clock

From January 1, 2027, new National Security System buys must support post-quantum algorithms. OMB wants high-value civilian systems on post-quantum key exchange by 2030. Attackers already use AI to turn patches into exploits. One inventory serves both clocks.

Ausper Technology · Nov 17, 2026 · Insight · Quantum and AI, 2027 · 9 min read · Updated Nov 18, 2026

The short version
  • No public evidence puts a code-breaking quantum computer in 2028 or 2029. Surveyed experts give 28 to 49 percent odds within 10 years, and paper estimates keep falling.
  • The deadlines are fixed: new National Security System buys must support CNSA 2.0 from January 1, 2027, and OMB M-26-15 sets December 31, 2030 for high-value civilian systems.
  • AI is shortening the gap from patch to exploit now. Mandiant estimates mean time to exploit at minus seven days, and 161 of 297 KEV entries in 2026 gave three days or fewer.

No public evidence says a quantum computer will break RSA or elliptic-curve cryptography by 2028 or 2029. The federal deadlines do not wait for one. From January 1, 2027, new acquisitions for National Security Systems must support NSA’s post-quantum algorithms, and OMB gives civilian high value assets and high impact systems until December 31, 2030 to move key exchange to post-quantum cryptography. The classical clock already runs faster than people: attackers use AI to turn new patches into working exploits. 2027 is where the two clocks meet. Both ask a defender where this runs and how fast you can change it.

The claim we tested

You hear it in every briefing this fall: public-key encryption gets broken in two or three years, and defenders working at human speed fall behind. We checked both halves. The second holds up. The first, as a date, does not. Our read: plan as if the window were short anyway.

The quantum clock: estimates fall, the machine is not here

The paper estimates keep dropping. In 2019, Gidney and Ekerå estimated that factoring a 2048-bit RSA key would take 20 million noisy qubits and eight hours. In May 2025, Gidney cut that to under a million qubits and under a week. A February 2026 design called Pinnacle put it near 100,000 physical qubits and about a month; it has not been peer reviewed (Quantum Computing Report). On October 29, Alice & Bob estimated 22,593 physical cat qubits and 3.14 days for RSA-2048, and 29,588 qubits and 61.8 minutes for a 256-bit elliptic curve (The Quantum Insider).

Fig. 1
On paper, breaking RSA-2048 fell from 20 million qubits to 22,593

Published estimates of the physical qubits needed to factor one RSA-2048 key. These are designs. None has been built.

PHYSICAL QUBITS, LOG SCALE10K100K1M10M100M2019202120232025202720 millionGidney and Ekerå, 2019 · 8 hoursunder 1 millionGidney, May 2025 · under a weekabout 100,000Pinnacle, Feb 2026 · about a month, not peer reviewed22,593Alice & Bob, Oct 2026 · 3.14 days, a theoretical projectionPHYSICAL QUBITS, LOG SCALE10K100K1M10M100M2019202120232025202720 millionGidney and Ekerå, 2019under 1 millionGidney, May 2025about 100,000Pinnacle, Feb 202622,593Alice & Bob, Oct 2026
Sources: Gidney, arXiv:2505.15917, which restates the 2019 estimate; Quantum Computing Report (Mar 6, 2026); The Quantum Insider (Oct 29, 2026). Each estimate assumes a different hardware design. "Under 1 million" is plotted at 1 million.
Data
PointValueNote
Gidney and Ekerå, 201920 million8 hours
Gidney, May 2025under 1 millionunder a week
Pinnacle, Feb 2026about 100,000about a month, not peer reviewed
Alice & Bob, Oct 202622,5933.14 days, a theoretical projection

Those are designs, not machines. The Quantum Insider calls the numbers theoretical projections, and Alice & Bob expects hardware and falling requirements to “converge in the early 2030s.” NIST’s final IR 8547, published in October, says no cryptanalytically relevant quantum computer exists today. In the Global Risk Institute survey published March 9, 26 experts put the chance of a machine that factors RSA-2048 in under 24 hours at 28 to 49 percent within 10 years. Within five years, the optimistic reading averages about 15 percent. Most respondents called that short-term risk low; 8 of 26 gave it 5 percent or more.

So two or three years is a tail risk, not a forecast. It is still the right planning stance. Adversaries can record encrypted traffic now and decrypt it later, which NIST calls one of the main reasons the transition is urgent. NIST also notes that moving from a new standard to full use has taken 10 to 20 years. Google moved its own migration target to 2029 in March, citing progress in hardware, error correction and factoring estimates.

The deadlines land on schedule

  • January 1, 2027. NSA’s CNSA 2.0 FAQ (version 2.1) requires new National Security System acquisitions to support CNSA 2.0, including ML-KEM-1024 and ML-DSA-87, unless an exception is noted. Equipment that cannot support it is to be phased out by December 31, 2030 (summary of the FAQ; NSA’s site refused our automated request).
  • 2027 to 2028. OMB M-26-15, issued June 24 under Executive Order 14412, schedules pilots and early migrations after a 2026 to 2027 inventory phase. Plans were due in 120 days.
  • January 2, 2030. Agencies must support TLS 1.3.
  • December 31, 2030, then 2031. High value assets, high impact systems and systems holding data that stays mission-sensitive in 2030 move key establishment to post-quantum algorithms by the end of 2030 and signatures by the end of 2031.
  • After 2030, after 2035. NIST deprecates quantum-vulnerable public-key algorithms after 2030 and disallows them under its standards after 2035.
Fig. 2
The first hard date is January 1, 2027. The civilian cliff is 2030.

Federal post-quantum milestones, in date order.

  1. Jun 24, 2026OMB sets the scheduleMemo M-26-15, under Executive Order 14412
  2. Oct 2026NIST finalizes its planIR 8547: which algorithms retire, and when
  3. Jan 1, 2027New national security buysmust support NSA's post-quantum algorithms
  4. Jan 2, 2030TLS 1.3 requiredCivilian agencies must support it
  5. Dec 31, 2030Key exchange movesHigh value and high impact civilian systems
  6. Dec 31, 2031Signatures moveThe same systems switch digital signatures
  7. 2035Everything elseThe rest migrate. NIST disallows the old algorithms.
Sources: OMB M-26-15; NIST IR 8547 (final, October 2026); NSA CNSA 2.0 FAQ v2.1, as summarized by PostQuantum.com.
Data
WhenMilestoneDetail
Jun 24, 2026OMB sets the scheduleMemo M-26-15, under Executive Order 14412
Oct 2026NIST finalizes its planIR 8547: which algorithms retire, and when
Jan 1, 2027New national security buysmust support NSA's post-quantum algorithms CNSA 2.0, unless NSA notes an exception
Jan 2, 2030TLS 1.3 requiredCivilian agencies must support it
Dec 31, 2030Key exchange movesHigh value and high impact civilian systems NSA phases out equipment that cannot support CNSA 2.0 by the same date; NIST deprecates RSA and elliptic curve after 2030
Dec 31, 2031Signatures moveThe same systems switch digital signatures
2035Everything elseThe rest migrate. NIST disallows the old algorithms. NIST disallows RSA and elliptic-curve algorithms after 2035

In August we wrote that quantum-ready is a deadline, not a physics problem. That holds. We cited M-23-02 and CNSA 2.0’s 2033 line then. We should also have cited M-26-15, which had set the civilian dates two months earlier.

The AI clock is already running

Mandiant’s M-Trends 2026 puts mean time to exploit at an estimated minus seven days: exploitation routinely starts before a patch exists. In May, Google’s threat intelligence group reported the first zero-day exploit it believes was developed with AI. On November 17 it described a framework that compares web application code before and after a patch, drafts an exploit and runs it, “without a human in the loop for the first time” (Help Net Security). It counted 22 zero-day exploits in August, against a 2025 average of 8 a month.

The skeptic’s data matters too. VulnCheck tracked 1,061 vulnerabilities attributed to AI-assisted discovery in the first half of 2026. Fourteen, or 1.3 percent, were confirmed exploited, about the overall rate. What moved was speed. The median time from CVE publication to listing as known exploited fell from 120 days in 2025 to 80, and 23.43 percent were exploited on or before the day the CVE was published. CISA has tightened to match: of 297 entries added to its KEV catalog in 2026 through November 13, 161 gave agencies three days or fewer.

Fig. 3
The classical clock already runs faster than a patch cycle
−7 daysMean time to exploit. On average, attacks start before the patch exists.Mandiant M-Trends 2026
161 of 297KEV entries added in 2026 that gave agencies three days or fewer.CISA KEV, through Nov 13
120 → 80Median days from CVE publication to known exploited, 2025 to first half of 2026.VulnCheck 1H-2026
22 vs 8Zero-day exploits Google counted in August, against a 2025 average of 8 a month.Google, via Help Net Security
Sources: Google Cloud, M-Trends 2026; CISA Known Exploited Vulnerabilities catalog, version 2026.11.13 (our count); VulnCheck, State of Exploitation 1H-2026; Help Net Security (Nov 17, 2026).
Data
NumberWhat it isSource
−7 daysMean time to exploit. On average, attacks start before the patch exists.Mandiant M-Trends 2026
161 of 297KEV entries added in 2026 that gave agencies three days or fewer.CISA KEV, through Nov 13
120 → 80Median days from CVE publication to known exploited, 2025 to first half of 2026.VulnCheck 1H-2026
22 vs 8Zero-day exploits Google counted in August, against a 2025 average of 8 a month.Google, via Help Net Security

Our read: AI is not yet finding more dangerous bugs. It is shrinking the time between a fix and a working attack, and that gap is where human speed loses.

Where the clocks meet

A three-day KEV and a 2030 key-exchange deadline fail on the same missing list. In October we showed the KEV window falling from 21 days to 3. On November 16 we wrote that AI tools now hold keys a patch cannot take back. Our November 17 2027 forecast asked who answers at 3 a.m. That person needs every component and algorithm in the boundary, with an owner.

Our forecasts

Our forecast: by December 31, 2027, no one will publicly demonstrate a quantum computer factoring an RSA-2048 key or recovering a 256-bit elliptic-curve key. We would bet on it, and we would still migrate as if we might lose.

Our forecast: in calendar 2027, more than half of the entries CISA adds to the KEV catalog will give federal agencies three days or fewer, as in 2026. If either forecast misses, we will say so here.

Our question: could you list today every system in your boundary that still uses RSA or elliptic-curve key exchange, and how long would that list take to build?

What we’d do first

  • Build the cryptographic inventory as a live feed. M-26-15 describes scanners, SBOM analysis and code scans filling a cryptographic bill of materials that feeds continuous-monitoring dashboards. Put it in the CMDB beside each system owner.
  • Rank by how long the data must stay secret. Anything still sensitive in 2030 goes first, because it can be recorded today.
  • Write CNSA 2.0 into requirements and subcontractor flow-down for any national security system delivered after January 1, 2027.
  • Put the migration in the ATO. SC-12 and SC-13 implementation statements, dated POA&M items, and continuous monitoring that shows progress.
  • Make the emergency change path fit three days, with named approvers and tested rollback.

Ausper is positioned to run both clocks from one inventory. Our capabilities include cryptographic inventory, post-quantum gap analysis against the NIST PQC standards and CNSA 2.0, and a migration roadmap written into your ATO artifacts, alongside continuous monitoring and vulnerability response around the clock. AccelSecure™ is the framework behind our 90 to 180 day ATO target. See our quantum work or compare notes with us.

Common questions

Will a quantum computer break RSA in the next two or three years?
No public evidence says so. NIST’s final IR 8547 (October 2026) says no cryptanalytically relevant quantum computer exists. In the Global Risk Institute’s survey of 26 experts, the optimistic average for a machine that factors RSA-2048 in under 24 hours within five years is about 15 percent, and most respondents called that risk low. Alice & Bob, whose October design needs 22,593 physical qubits, expects hardware to catch up in the early 2030s. Data recorded today can still be decrypted later, which is why the migration cannot wait.
What is due on January 1, 2027?
NSA’s CNSA 2.0 FAQ (version 2.1, December 2024) requires new acquisitions for National Security Systems to support CNSA 2.0 algorithms, including ML-KEM-1024 and ML-DSA-87, unless an exception is noted. It applies to national security systems; civilian agencies follow OMB M-26-15 and NIST IR 8547.
What does OMB M-26-15 require of civilian agencies?
A PQC migration plan to OMB and ONCD within 120 days of the June 24, 2026 memo, then a phased migration: inventory and strategy in 2026 to 2027, pilots in 2027 to 2028, post-quantum key establishment for high value assets and high impact systems by December 31, 2030, and signatures in 2031. Agencies must support TLS 1.3 by January 2, 2030, and align plans with NIST IR 8547.
Is AI making exploitation faster?
The timing data says yes, though AI-found bugs are not yet exploited more often. VulnCheck found 1.3 percent of 1,061 AI-attributed vulnerabilities confirmed exploited, about the overall rate, while the median time to exploitation evidence fell from 120 to 80 days. Mandiant estimates mean time to exploit at minus seven days, and Google reported in November an AI framework that turns patches into exploits with no human in the loop.

Sources

  1. Craig Gidney, How to factor 2048 bit RSA integers with less than a million noisy qubits (arXiv:2505.15917, May 2025)Abstract: under a million noisy qubits, under a week; recalls the 2019 Gidney and Ekerå estimate of 20 million qubits and eight hours. Accessed Nov 17, 2026.
  2. Quantum Computing Report, Shor, QLDPC Codes, and the Compression of RSA-2048 Resource Estimates (Part I) (Mar 6, 2026)Pinnacle Architecture, February 2026: about 100,000 physical qubits, about one month, not yet peer reviewed. Accessed Nov 17, 2026.
  3. The Quantum Insider, Alice & Bob Estimates 22,593 Qubits to Break RSA-2048 (press release, Oct 29, 2026)RSA-2048: 22,593 physical cat qubits, 3.14 days; ECDLP-256: 29,588 qubits, 61.8 minutes; theoretical projections; convergence expected in the early 2030s. Accessed Nov 17, 2026.
  4. Global Risk Institute and evolutionQ, Quantum Threat Timeline Report 2025 (Mosca and Piani, published Mar 9, 2026)26 experts; 28 to 49 percent within 10 years; full report PDF read for the five-year figures (about 15 percent optimistic average; 8 of 26 at 5 percent or more). Accessed Nov 17, 2026.
  5. Google, Quantum frontiers may be closer than they appear (Mar 25, 2026)Google's 2029 timeline for its own post-quantum migration. Accessed Nov 17, 2026.
  6. NIST IR 8547, Transition to Post-Quantum Cryptography Standards (final, October 2026)PDF read: no existing CRQC; harvest now, decrypt later; 10 to 20 years for past transitions; deprecated after 2030, disallowed after 2035; restates EO 14412 and M-26-15 dates and the TLS 1.3 deadline. Accessed Nov 17, 2026.
  7. OMB M-26-15, Execution of the Migration to Post-Quantum Cryptography (June 24, 2026)120-day plans; phases 2026 to 2035; prioritized systems; automated inventory and CBOM. Accessed Nov 17, 2026.
  8. GovInfo, Executive Order 14412, Securing the Nation Against Advanced Cryptographic Attacks (June 22, 2026)Accessed Nov 17, 2026.
  9. PostQuantum.com, NSA Updates CNSA 2.0 Guidance After NIST Finalizes Post-Quantum Standards (Dec 29, 2024)Summary of NSA's CNSA 2.0 FAQ version 2.1: the January 1, 2027 acquisition requirement, ML-KEM-1024 and ML-DSA-87, phase-out by December 31, 2030. media.defense.gov refused automated requests. Accessed Nov 17, 2026.
  10. Google Cloud, M-Trends 2026: Data, Insights, and Strategies From the FrontlinesMean time to exploit an estimated -7 days (2025 investigations). Accessed Nov 17, 2026.
  11. Google Threat Intelligence Group, GTIG AI Threat Tracker on AI for vulnerability exploitation and initial access (May 11, 2026)First zero-day exploit GTIG believes was developed with AI. Accessed Nov 17, 2026.
  12. Help Net Security, Google finds AI agent writing and launching exploits with no human in the loop (Nov 17, 2026)GTIG and Mandiant: exploit_agent framework; 22 zero-day exploits in August 2026 against a 2025 average of 8 a month. Accessed Nov 17, 2026.
  13. VulnCheck, State of Exploitation 1H-2026 (July 28, 2026)1,061 AI-attributed vulnerabilities, 14 exploited; median CVE-to-KEV 120 to 80 days; 23.43 percent exploited on or before CVE publication. Accessed Nov 17, 2026.
  14. CISA, Known Exploited Vulnerabilities catalog, version 2026.11.13 (JSON, public mirror cisagov/kev-data)161 of 297 entries added in 2026 gave three days or fewer; computed by make_first_hour_2027.py. Accessed Nov 17, 2026.

Updates

  • Nov 18, 2026. Added three figures: the falling qubit estimates, the deadline timeline and the exploit-clock numbers. No text or data changed.

Get Insights by email

Thanks for subscribing.

Share on LinkedInFollow AusperSubscribe by RSS

Related reading

Put this to work

Working on this now?

Tell us where your program stands and we'll tell you what we would do next.

Talk to Ausper