Quantum Entanglement Explained: What It Is and What It Is Not

Quantum entanglement is a shared state of two or more particles in which the measurement results are correlated more strongly than any theory of preset local properties allows, and Bell tests since 1972 have confirmed it, culminating in the 2022 Nobel Prize in Physics. It is not a hidden wire between particles: neither side can control its own outcome, so entanglement cannot send a message, a bit or a signal faster than light. What it can do is serve as a resource, for teleporting quantum states, for quantum computing, for sensing and for future quantum networks.

This post explains the idea without hype, walks through the history from Einstein, Podolsky and Rosen in 1935 to the loophole free tests of 2015, shows why the no communication theorem holds, and gives you a short Python simulation of a Bell test that you can run and check yourself. Every date below was checked against a primary source: the Nobel Committee's own background papers, the original journal records, NIST, the US Department of Energy and arXiv.

Key takeaways

• An entangled pair has a definite joint state but no definite state for each part, and its correlations beat any local instruction set.
• Einstein, Podolsky and Rosen argued in May 1935 that quantum mechanics is incomplete. Bell showed in 1964 that local hidden variables make testable, different predictions.
• Freedman and Clauser saw the first clear violation in 1972. Aspect's 1982 experiment switched settings while the photons were in flight. In 2015 groups in Delft, Vienna and Boulder closed the main loopholes together.
• The 2022 Nobel Prize in Physics went to Alain Aspect, John Clauser and Anton Zeilinger, announced on 4 October 2022.
• Entanglement cannot transmit information. Each side's results look like coin flips whatever the other side does; the correlation appears only when records are compared over an ordinary channel.
• Quantum teleportation moves an unknown quantum state, not matter, and it needs two classical bits per qubit sent at light speed or slower.
• Entanglement cannot be amplified. Quantum repeaters, built on swapping, purification and memory, are the planned fix, and the US DOE's July 2020 blueprint maps a route to a quantum internet.

A dated timeline of entanglement

One date wrinkle: the APS record lists Bell's paper as published 1 November 1964 but received 4 November 1964, and the 1969 CHSH paper cites it as 1965. The Nobel Committee and NIST both say 1964.

What is quantum entanglement, in plain terms?

The Nobel Committee's popular background starts with two balls, one black and one white, sent in opposite directions. Catch white and you know the other is black; nothing travelled, because the colours were fixed at the start. In quantum mechanics, it says, the balls behave as if both are grey until someone looks, and then one randomly turns white and the other black. The question that took fifty years to settle was whether each ball secretly carried a hidden label saying which colour to show.

More precisely: some two qubit states factor into "a state for A" times "a state for B". Those are product states. Others cannot be factored, and those are entangled. The textbook example is the singlet:

|psi-> = ( |0>_A |1>_B  -  |1>_A |0>_B ) / sqrt(2)

The pair has a definite joint state, yet each qubit alone is maximally uncertain. Measure A along any axis and each outcome comes up half the time; measure B along the same axis and it always disagrees with A. Along different axes, the agreement rate follows the cosine of the angle between them. Hidden labels can fake the anticorrelation along one axis, but not the full cosine pattern across every pair of angles. Bell turned that into an inequality.

NIST's 2026 network release gives the one line version: entangled objects "cannot be described independently even if they are far apart." Entanglement is about description and correlation, not a force acting between two places.

Why did Einstein, Podolsky and Rosen call quantum mechanics incomplete?

The EPR paper, received by Physical Review on 25 March 1935 and published on 15 May 1935, starts from a criterion in its abstract: if you can predict a quantity with certainty without disturbing the system, there is an element of reality corresponding to it. Take two systems that interacted and are now apart. Measuring position on the first predicts the second's position; measuring momentum predicts its momentum. Since either choice could be made far away, EPR conclude the second system had both values all along. Quantum mechanics does not assign both at once, so, they argue, it is incomplete.

The hidden assumption is locality. Bell's paper quotes Einstein saying that the real situation of the second system "is independent of what is done" with the first.

EPR did not claim quantum predictions were wrong, only that extra variables were missing. For almost thirty years both views predicted the same results for any experiment people had in mind, so the question looked philosophical. Bell changed that.

Schrödinger answered the same year in a Cambridge paper on separated systems; the Nobel popular background notes he called entanglement quantum mechanics' most important trait.

What did John Bell prove in 1964?

Bell asked whether any local hidden variable theory could reproduce all of quantum mechanics' predictions, and showed it cannot. Assume that:

1. Each particle carries instructions, set at the source, that fix its result for every possible measurement setting (realism, in Bell's technical sense).
2. The result at one detector cannot depend on the setting chosen at the other, distant detector (locality).
3. The experimenters choose their settings freely, independently of the instructions in the particles (sometimes called measurement independence or freedom of choice).

From those assumptions alone, correlations must obey a bound, a Bell inequality. Quantum mechanics predicts entangled pairs exceed it at certain angles, so at least one assumption must fail.

The CHSH version that experiments can test

Bell's original inequality assumed ideal detectors. In 1969 Clauser, Horne, Shimony and Holt published a version that real equipment could test. Their abstract says it generalises Bell's theorem "so as to apply to realizable experiments".

Alice picks setting a1 or a2, Bob b1 or b2, and each records plus one or minus one. For each pair of settings, the correlation E is the average product of the results. Combine four:

S = | E(a1,b1) - E(a1,b2) + E(a2,b1) + E(a2,b2) |

Any local hidden variable model gives S at most 2. Quantum mechanics, with the singlet and the right angles (Alice at 0 and 90 degrees in the spin picture, Bob at 45 and 135 degrees), gives S equal to 2 times the square root of 2, about 2.83. Every Bell test looks for that gap, and the walkthrough below reproduces both numbers.

How did Bell tests confirm entanglement?

Freedman and Clauser, 1972

According to the Nobel background, Clauser and doctoral student Stuart Freedman rebuilt an earlier calcium cascade experiment at Berkeley so it could test the CHSH inequality, and recorded data for about 200 hours. Their 1972 result, published in PRL 28, 938, violated the inequality by about six standard deviations and agreed with quantum mechanics.

Aspect, 1981 to 1982

Clauser's polarizers were fixed during each run, so information about the settings could in principle have leaked. Alain Aspect, with Philippe Grangier, Gérard Roger and Jean Dalibard, ran three experiments; the second gave the largest violation at the time, tens of standard deviations, and the third switched settings while the photons were in flight. With six metres from source to each polarizer, switching had to happen within a few billionths of a second.

Zeilinger and the move to applications

Anton Zeilinger's group made entangled photons with a laser and a nonlinear crystal and chose settings with random numbers; one experiment, the Nobel popular background notes, used light from distant galaxies to pick them. His group also turned entanglement into tools: teleportation in 1997, entanglement swapping in 1998, and Ekert style key distribution over 144 km between two Canary Islands in 2006 (S = 2.508 plus or minus 0.037).

The loopholes, and closing them in 2015

Every real Bell test makes some extra assumption that a determined skeptic could exploit. The main ones are below.

Before 2015, experiments closed one loophole or another but not the main ones together; trapped ions first closed detection without closing locality. In 2015 three experiments did both.

NIST's 10 November 2015 release puts the Boulder odds at about 1 in 170 million for local realism, beyond the five sigma discovery threshold. NIST physicist Krister Shalm framed it carefully: "You can't prove quantum mechanics," but local realism is incompatible with the result. Bell tests rule out the local hidden variable theories EPR hoped for; they do not prove quantum mechanics in some absolute sense.

Why did the 2022 Nobel Prize in Physics go to entanglement?

On 4 October 2022 the Royal Swedish Academy of Sciences awarded the prize to Alain Aspect (Institut d'Optique and École Polytechnique), John F. Clauser (J.F. Clauser and Associates, California) and Anton Zeilinger (University of Vienna), "for experiments with entangled photons, establishing the violation of Bell inequalities" and, the citation continues, pioneering quantum information science. The 10 million Swedish kronor was shared equally.

The press release credits Clauser with the first practical test, Aspect with closing an important loophole and Zeilinger with turning entangled states into protocols such as teleportation.

The scientific background adds that the main importance of the 2015 to 2017 loophole free tests is not to confirm quantum mechanics again but to enable more secure key distribution: the question becomes whether an eavesdropper, not nature, could fake a violation. In 2022 three groups used loophole free Bell tests for device independent key distribution, secure even if the attacker built the hardware.

Why can't entanglement send information faster than light?

This is the single most important thing to get right, and the most common thing popular accounts get wrong.

The short answer

Whatever Alice does to her particle, Bob's own results, on their own, have exactly the same statistics. Alice cannot choose her outcome, only her setting, and her setting does not change the probabilities Bob sees. The correlation exists only in the joint record, and to build the joint record Alice and Bob must bring their notes together over an ordinary channel, at light speed or slower.

The slightly longer answer

Everything Bob can learn from his qubit alone is captured by its reduced state. For the singlet that is the maximally mixed state: equal odds along every axis. If Alice measures along any axis, each of her outcomes leaves Bob's qubit in a definite state, but Bob does not know which; averaged over her outcomes, his state is still maximally mixed. Nothing Alice does locally changes Bob's reduced state. That is the no communication theorem, or no signalling.

Bob might hope to cheat by copying his qubit many times and measuring the copies to work out which state it is in. The no cloning theorem, proved by Wootters and Zurek in 1982, closes that door. Their Nature abstract shows "the linearity of quantum mechanics forbids such replication" for every quantum system. If cloning were possible, faster than light signalling would follow, so the two results protect each other.

The Bell test numbers show the same thing from the other side. In the simulation below, Bob's probability of getting plus one is 0.5000 whether Alice measures at 0 degrees or 90 degrees. Only when the two lists are lined up does the 2.83 appear.

Peres and Terno's review of quantum information and relativity sums it up: special relativity severely restricts information transfer between distant systems, and entanglement respects that.

So what is "spooky" about it?

The correlations themselves. EPR hoped instructions set at the source could explain them; Bell tests show they cannot, so the joint state is not reducible to local parts. But this nonlocality cannot be used to signal, which is why entanglement and relativity coexist.

How does quantum teleportation work, and what does it actually move?

Teleportation, proposed by Bennett and colleagues in PRL on 29 March 1993, moves an unknown quantum state from one particle to another without sending the particle. Their abstract describes splitting the state into "purely classical information" and purely nonclassical EPR correlations.

The steps:

1. Alice and Bob share an entangled pair in advance. Alice holds one half, Bob the other.
2. Alice has a third qubit in an unknown state she wants to send. She performs a joint measurement (a Bell state measurement) on that qubit and her half of the pair.
3. The measurement gives one of four outcomes, two classical bits. It also destroys the original state on Alice's side.
4. Alice sends the two bits to Bob over an ordinary channel.
5. Bob applies one of four fixed corrections to his half, chosen by the two bits. His qubit is now in the original unknown state.

What moves is the quantum state. No matter is beamed anywhere, and the original is gone, so no cloning holds. Until Bob gets the two bits, his qubit is maximally mixed, so teleportation cannot outrun light either.

The Nobel popular background adds that teleportation is the only way to move quantum information without losing part of it, since you cannot measure every property of an unknown state and rebuild it.

Teleportation milestones

• 1997: first experiments, by Zeilinger's group (Bouwmeester et al., published in Nature on 1 December 1997) and by Francesco De Martini's group.
• 2017: single photon qubits teleported from a ground observatory to the Micius satellite over up to 1,400 km; earlier ground links were limited to about 100 km by photon loss.
• 2021: Delft teleported qubits between two nodes that were not directly connected, via a third node that performed entanglement swapping and stored the result in memory, with fidelity above the classical bound.

How is entanglement used as a resource?

Once Bell tests settled the physics, entanglement became something to produce, distribute, store and consume. Here are the main uses, from primary sources and without forecasting.

Computing

Entangling operations between qubits are a core ingredient of quantum algorithms. NIST's quantum computing explainer describes entangled qubits sharing one quantum state with no independent parts. The catch is fragility: entanglement leaks into the environment, which is why error correction matters so much. For the cryptography angle, see our explainer on a cryptographically relevant quantum computer.

Sensing

Quantum sensing, in the definition used by Degen, Reinhard and Cappellaro's 2017 review in Reviews of Modern Physics, means using a quantum system, quantum properties or quantum phenomena to measure a physical quantity. The review's historical examples, atomic clocks and magnetometers based on SQUIDs or atomic vapours, work without entanglement. Entanglement and the closely related squeezed states help when the limit is quantum noise itself.

The largest working example is gravitational wave detection. NIST's sensing explainer describes how LIGO in 2019 installed crystals that make entangled photon pairs and squeeze the vacuum light entering the detector, reducing phase uncertainty, and notes that LIGO, Virgo and KAGRA are among the few sensors to gain precision this way. Ideas such as entangled telescope or seismic networks, mentioned in NIST's 2026 release, remain research goals.

Networking and key distribution

In Artur Ekert's 1991 protocol, Alice and Bob measure entangled pairs, publicly compare a subset to run a Bell test, and, if the violation is there, turn the remaining correlated results into a secret key. Device independent versions, shown in 2022, extend the guarantee to untrusted hardware. QKD shares keys, not messages; the message still travels over a classical link. We cover it in quantum key distribution explained and PQC vs QKD: what each does.

What are quantum repeaters and the quantum internet?

The distance problem

Light fades in fibre. The Nobel background says that on average every second photon is lost in 10 km of fibre (actual loss depends on fibre and wavelength; the Micius paper cites 0.16 dB per km for the best commercial fibre). Classical networks use amplifiers, but an amplifier has to capture and measure the light, which breaks entanglement, and copying the state is forbidden by no cloning. Loss grows exponentially with distance, so there are two answers.

Satellites

Most of a satellite link runs through empty space. The Micius team delivered entangled pairs to ground stations 1,203 km apart, through downlinks totalling 1,600 to 2,400 km, with a Bell violation of 2.37 plus or minus 0.09, and estimated the link efficiency at more than 12 orders of magnitude above the best commercial fibre. The satellite later served as a trusted relay for a Beijing to Vienna key, per the Nobel background; a trusted relay knows the key, a different security model from end to end entanglement.

Repeaters: swapping, purification and memory

A quantum repeater does not amplify. It creates entanglement on short segments, then uses entanglement swapping at middle nodes to join them into one long pair; the DOE blueprint describes the repeater as letting entanglement "hop" an extra interval. Two more pieces are needed. Purification distils fewer, better pairs from noisy ones; Briegel, Dür, Cirac and Zoller's 1998 scheme nests it with swapping, tolerating percent level errors with polynomial time overhead. Quantum memory holds each segment's entanglement until its neighbours succeed, which the Nobel background lists as a requirement.

The DOE blueprint

On 23 July 2020, at the University of Chicago, the US Department of Energy unveiled a strategy for a national quantum internet, following a 5 and 6 February 2020 workshop in New York with national laboratories, universities and industry. The release names DOE's 17 national laboratories as the backbone and mentions a 52 mile entanglement loop in the Chicago suburbs, due to join Fermilab in an 80 mile, three node testbed.

The report sets out four priority research directions and five milestones:

The report notes estimates that more than 100 repeaters could be needed for a transatlantic link, and compares a proposed quantum network stack with the TCP/IP model: the network layer builds long distance entanglement by swapping, and the transport layer moves qubits deterministically by teleportation.

The DOE release speaks of "virtually unhackable networks". Treat that as an aim: a deployed link's security depends on implementation, trusted nodes and authentication, which still uses classical cryptography. Progress comes in measurable steps; in August 2026 NIST reported entangled photons surviving 62 km of commercial aerial fibre between Gaithersburg and College Park, Maryland, through temperature swings and wind.

Correlation, entanglement and signalling compared

Step by step: simulate a Bell test and check no signalling yourself

You do not need a lab to see the numbers. The script below computes the quantum predictions for the singlet, brute forces the best local strategy, prints Bob's marginals for each of Alice's settings, and samples simulated pairs. It simulates the mathematics, so it cannot rule out hidden variables itself; real Bell tests do that. It shows what quantum theory predicts and why that prediction carries no message.

1. Create an isolated environment and pin the dependency

Use a virtual environment and pin the one dependency so the result is reproducible.

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install "numpy==2.0.2"
python -m pip freeze > requirements.txt

For stronger guarantees, add hashes to the pinned requirements and install with pip install --require-hashes -r requirements.txt, so a changed artifact fails the install.

2. Save the script

Save this as chsh_demo.py:

import numpy as np

rng = np.random.default_rng(seed=7)

# Singlet state (|01> - |10>)/sqrt(2)
psi = np.array([0, 1, -1, 0]) / np.sqrt(2)

def projector(angle, outcome):
    # Spin measurement along an angle in the x-z plane; outcome +1 or -1
    if outcome == 1:
        v = np.array([np.cos(angle / 2), np.sin(angle / 2)])
    else:
        v = np.array([-np.sin(angle / 2), np.cos(angle / 2)])
    return np.outer(v, v)

def joint_probs(a, b):
    p = {}
    for x in (1, -1):
        for y in (1, -1):
            P = np.kron(projector(a, x), projector(b, y))
            p[(x, y)] = float(psi @ P @ psi)
    return p

def correlation(a, b):
    p = joint_probs(a, b)
    return sum(x * y * q for (x, y), q in p.items())

a1, a2 = 0.0, np.pi / 2
b1, b2 = np.pi / 4, 3 * np.pi / 4

S_quantum = abs(correlation(a1, b1) - correlation(a1, b2)
                + correlation(a2, b1) + correlation(a2, b2))
print(f"Quantum CHSH value S = {S_quantum:.4f}  (2*sqrt(2) = {2*np.sqrt(2):.4f})")

# No signalling check: Bob's marginal for each of Alice's settings
for a in (a1, a2):
    p = joint_probs(a, b1)
    bob_plus = p[(1, 1)] + p[(-1, 1)]
    print(f"Alice setting {a:.3f} rad -> P(Bob gets +1) = {bob_plus:.4f}")

# Best deterministic local hidden variable strategy: try all 16
best = 0
for A1 in (1, -1):
    for A2 in (1, -1):
        for B1 in (1, -1):
            for B2 in (1, -1):
                best = max(best, abs(A1*B1 - A1*B2 + A2*B1 + A2*B2))
print(f"Best local deterministic strategy S = {best}")

# Sampled run, 100000 pairs per setting combination
def sample(a, b, n=100_000):
    p = joint_probs(a, b)
    keys = list(p)
    idx = rng.choice(4, size=n, p=[p[k] for k in keys])
    return np.mean([keys[i][0] * keys[i][1] for i in idx])

S_sampled = abs(sample(a1, b1) - sample(a1, b2) + sample(a2, b1) + sample(a2, b2))
print(f"Sampled S from 400000 simulated pairs = {S_sampled:.3f}")

3. Run it and read the output

python chsh_demo.py

On our machine with NumPy 2.0.2 it printed:

Quantum CHSH value S = 2.8284  (2*sqrt(2) = 2.8284)
Alice setting 0.000 rad -> P(Bob gets +1) = 0.5000
Alice setting 1.571 rad -> P(Bob gets +1) = 0.5000
Best local deterministic strategy S = 2
Sampled S from 400000 simulated pairs = 2.838

Line by line:

• Line one is the quantum CHSH value, 2.83, above 2.
• Lines two and three are the no signalling check: Bob's chance of plus one is one half whichever setting Alice uses.
• Line four tries every set of preset answers, which covers every deterministic local model (random mixtures cannot beat the best one). The maximum is 2.
• Line five samples finite data; noise moves it slightly off 2.83, which is why real experiments report error bars.

4. Try breaking it

Replace psi with a product state such as np.array([1, 0, 0, 0]) and S never exceeds 2, while Bob's marginals still ignore Alice's setting. The violation comes from entanglement; the absence of signalling holds either way.

Myth vs fact

What this means for security teams

Entanglement on its own does not break or weaken any cryptography you run today. The threat people worry about is a large, error corrected quantum computer running Shor's algorithm against RSA and elliptic curve keys, which is a computing question, not an entanglement trick. The defensive response is post quantum cryptography, which runs on ordinary hardware; see our post quantum cryptography overview and when a quantum computer could break encryption.

Entanglement based key distribution is a separate, hardware based approach with limits of distance, trusted relays and classical authentication. If a pitch says entanglement makes a link instant or unhackable, test it against the facts above.

Where Safeguard fits, honestly

Safeguard does not do quantum anything. It does not build quantum networks, distribute keys, simulate qubits or assess quantum hardware, and nothing in this post describes a Safeguard feature. The overlap is narrow and ordinary: quantum research and teaching code is software, usually Python or C++ with deep dependency trees, and it lives in repositories and container images like any other code. Safeguard scans repositories, container images and SBOM files, matches components against vulnerability databases and shows which projects carry a given component and version. If your lab or team builds simulation notebooks, control software or analysis pipelines, safeguard scan --dir . on the repository or safeguard scan --image <name> on the image tells you what is inside and what needs attention.

Common mistakes

These are the common misconceptions we see when entanglement comes up in technical conversations.

• Saying measurement on one side "changes" the other in a usable way. Measurement updates the joint description, but nothing observable changes for the other party until records are compared.
• Treating teleportation as instant. It is gated by a classical message. Until Bob gets Alice's two bits, his qubit is useless noise.
• Calling 1972 or 1982 the final word. Those experiments left loopholes open. The strong result, with locality and detection closed together, came in 2015.
• Assuming the quantum internet is a faster internet. The DOE blueprint describes new functions, such as distributing entanglement and linking quantum processors and sensors, not more bandwidth for web traffic.
• Reading "trusted relay" satellite links as end to end entanglement. A trusted relay knows the key. That is a different trust model.
• Running copied simulation code without pinning dependencies. Teaching notebooks often install unpinned packages. Pin versions and isolate environments so the result you publish is the result others reproduce.

FAQ

What is quantum entanglement in simple terms?

Two or more particles share one quantum state that cannot be split into separate states for each. Measuring them gives results that are correlated more strongly than any preset local instructions could explain.

Can quantum entanglement be used to communicate faster than light?

No. Each side's own results look random whatever the other side does, so no message can be encoded. The correlation only shows up after comparing results over a normal channel.

What did Einstein mean by spooky action at a distance?

He was objecting to the idea that measuring one particle seems to fix the state of a distant one. NIST and the Nobel Committee both use the phrase; Bell tests show the correlations are real but not usable for signalling.

What did the 2022 Nobel Prize in Physics recognise?

Experiments by Alain Aspect, John Clauser and Anton Zeilinger with entangled photons that established the violation of Bell inequalities and opened quantum information science. It was announced on 4 October 2022.

Does quantum teleportation move matter?

No. It transfers the quantum state of one particle onto another that is already at the destination, destroys the original state and needs a classical message.

What is a quantum repeater?

A node that extends entanglement over long distances by creating entanglement on short segments, purifying it and joining segments with entanglement swapping, using quantum memory to wait for each step to succeed.

What is the DOE quantum internet blueprint?

A US Department of Energy report released on 23 July 2020 that sets out four priority research directions and five milestones toward a national quantum internet, from secure protocols over fibre to interstate repeater networks.

Does entanglement break encryption?

Not by itself. The cryptographic threat comes from large quantum computers running Shor's algorithm. Entanglement can also be used defensively, in key distribution protocols.

Fix it without waiting for upstream

Nothing in quantum physics changes how software gets patched, and this section is about the ordinary code around the research, not the physics. Simulation toolkits, lab control software and analysis notebooks pull in open source dependencies like any other project, and a vulnerable version in that stack is a normal supply chain problem with normal fixes.

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation covers eight categories, including dependencies, licenses, security posture and findings. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds. The safeguard-sh/gold-scan-action GitHub Action can fail a build on a known exploited vulnerability, on any known vulnerability, or never, so a version you have just removed cannot quietly come back in next week's merge.

3. Gold Registry: start from components that are already fixed

registry.safeguard.sh provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs, zero malware and a completed license check, with SLSA L3 or higher provenance, a Sigstore signature and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. This fork and patch flow is in early access today, so ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

To try it on your own machine, download Safeguard from https://safeguard.sh/download and see the CLI documentation at https://docs.safeguard.sh/docs/cli .

See how Safeguard would help

The first demo is a real Safeguard CLI session. Asked to fix a lodash CVE, it reads the finding and advisory, asks before each command, edits package.json to lodash 4.17.21, runs npm install and npm test with approval, and pushes nothing. The same approve each step flow applies to upgrading a pinned dependency in a Python simulation or analysis repository.

Click through it at your own pace, or open it full screen.

The second demo sets up Safeguard's local runner on your own machine, approves it with a device code, and runs a scan from the terminal. On a research workstation, the same flow scans a notebook repository or a lab image without sending the code anywhere first.

Click through it at your own pace, or open it full screen.

The third demo connects Claude to the Safeguard MCP server with OAuth and lets you choose which tools Claude can use. That lets you ask which of your projects carry a given package version while controlling exactly what the assistant can reach.

Click through it at your own pace, or open it full screen.

Quick check

Sources

• Nobel Prize in Physics 2022: press release, 4 October 2022
• Nobel Prize in Physics 2022: popular science background, "How entanglement has become a powerful tool" (PDF)
• Nobel Prize in Physics 2022: scientific background (PDF)
• Einstein, Podolsky and Rosen, "Can Quantum Mechanical Description of Physical Reality Be Considered Complete?", Phys. Rev. 47, 777 (1935)
• Schrödinger, "Discussion of Probability Relations between Separated Systems", Proc. Camb. Phil. Soc. 31, 555 (1935)
• Bell, "On the Einstein Podolsky Rosen paradox", Physics Physique Fizika 1, 195 (1964)
• Clauser, Horne, Shimony and Holt, "Proposed Experiment to Test Local Hidden Variable Theories", PRL 23, 880 (1969)
• Wootters and Zurek, "A single quantum cannot be cloned", Nature 299, 802 (1982)
• Bennett et al., "Teleporting an unknown quantum state via dual classical and Einstein Podolsky Rosen channels", PRL 70, 1895 (1993)
• Bouwmeester et al., "Experimental quantum teleportation", Nature 390, 575 (1997)
• Briegel, Dür, Cirac and Zoller, "Quantum Repeaters: The Role of Imperfect Local Operations in Quantum Communication", PRL 81, 5932 (1998)
• Hensen et al., "Experimental loophole free violation of a Bell inequality using entangled electron spins separated by 1.3 km", arXiv:1508.05949
• Giustina et al., "Significant loophole free test of Bell's theorem with entangled photons", arXiv:1511.03190
• Shalm et al., "A strong loophole free test of local realism", arXiv:1511.03189
• NIST: "NIST Team Proves 'Spooky Action at a Distance' is Really Real", 10 November 2015
• Yin et al., "Satellite Based Entanglement Distribution Over 1200 kilometers", arXiv:1707.01339 (Science 356, 1140, 2017)
• Ren et al., "Ground to satellite quantum teleportation", arXiv:1707.00934
• Hermans et al., "Qubit teleportation between non neighboring nodes in a quantum network", arXiv:2110.11373
• Peres and Terno, "Quantum information and relativity theory", arXiv:quant-ph/0212023
• Degen, Reinhard and Cappellaro, "Quantum sensing", arXiv:1611.02427 (Rev. Mod. Phys. 89, 035002, 2017)
• US Department of Energy: "U.S. Department of Energy Unveils Blueprint for the Quantum Internet", 23 July 2020
• US Department of Energy: Report of the DOE Quantum Internet Blueprint Workshop, "From Long distance Entanglement to Building a Nationwide Quantum Internet" (PDF)
• NIST: "Spooky Particles Transit DC Suburbs in Step Toward Quantum Network", 5 August 2026
• NIST: Quantum computing explained
• NIST: Squeezed light for detecting gravitational waves

Last reviewed: 23 October 2026.