Aiseptor

Aiseptor is an exam integrity tool for online tests and interviews that works on the candidate's device and network rather than through a camera. Its homepage calls it "a network-layer security platform that blocks AI cheating at the device and network layer, not the browser." The problem it targets is a specific one: invisible AI overlays that hide from screen sharing, remote-access tools, and language models running locally on the candidate's own computer. The company's line is "Exam security without the surveillance model": no webcam, no screen recordings, no persistent extension.

According to the About section, Aiseptor Inc. was founded in 2025 by Akshay Aggarwal, Divya Bhanushali and Sanjay Ram, and won the 2026 Cornell Tech Startup Award. It is an early-stage product, and the site is open about that: the homepage counts 3 signed pilots with assessment platforms and two live runs at Cornell University with 100+ candidates each.

How it works

When an exam starts, Aiseptor deploys what it calls "an ephemeral security enclave" on the candidate's machine, on any device, in 30 seconds, and the homepage's section for hiring teams describes it as "30-second deploy, 10-second destruct." The homepage stresses that there are no kernel drivers and no persistent install, and that enforcement runs on unmanaged devices. The exam itself can open in a normal browser while Aiseptor enforces the machine boundary.

The detection approach is the core of the pitch. Rather than keeping a list of banned tools by name, Aiseptor looks for the technique those tools rely on. Every overlay that hides from screen share has to set the same screen-capture exclusion flag at the operating system level (the site names it as WDA_EXCLUDEFROMCAPTURE), so Aiseptor flags that, along with install-path, Electron renderer-tree and GPU-memory signals. The company argues this catches renamed builds, open-source forks and tools it has never seen, as well as named ones such as Cluely. For language models running in local memory, which make no network request and show nothing on screen, it scans for process, model-file and GPU-memory signatures, naming Ollama and LM Studio as examples.

The homepage frames exam security as three layers. Above sits the physical and behavioral layer: webcams, gaze tracking and human proctors. Below sits the application and browser layer, where lockdown browsers control a single window. Aiseptor claims the middle layer, the device and the network, which it says is where invisible cheating now happens: hidden overlays, remote-access tools, local LLMs and second-device pivots. Enforcement is scoped to the session only, there is no browser extension, and the enclave tears itself down when the session ends.

What it records is narrow. The site describes the log as network-reachability results and device-integrity signals: which hosts a device could reach and whether an exclusion flag was set, not payloads and not footage. The audit record is "cryptographically signed metadata, not footage of your candidates."

What the proctor sees

The console at dashboard.aiseptor.com shows each session's posture. In the homepage example, a candidate launches a screen-capture tool mid-exam; Aiseptor blocks the path, the session flips from clean to incident, and the proctor receives the process name, the candidate and a remediation step.

Aiseptor separates enforcement from judgment on purpose. It enforces, and the Aegis report shows which signals fired and which did not; what a signal means for a candidate is left to the customer's own platform. The homepage claims "Zero bypasses. Zero false positives," explaining that because Aiseptor closes the path instead of scoring behavior, "a session is either enforced or it is not." It reports a 0% bypass rate across 25+ attack vectors attempted in live pilot runs, and says that in one design partner's own adversarial test it "covered 100% of their threat model." These are the company's own figures from pilots.

Secure Browser

Alongside the network layer, Aiseptor offers Aiseptor Secure Browser, a lockdown browser for Mac and Windows that is currently in beta. The homepage says it now covers the application and browser layer as well as the device and network layer.

Who uses it, and how

The homepage describes three ways to deploy:

• Assessment platforms embed Aiseptor beneath their own product through "One REST API," with a developer quickstart on the site.
• Engineering and hiring teams use it for technical interviews, blocking overlays, remote-access tools and local LLMs during the call.
• Certification bodies use it to protect at-home exams without a physical test center.

The company positions Aiseptor as a layer that sits under whatever proctoring or lockdown tool a customer already has, rather than a full replacement.

Privacy, data and trust

Post-exam records default to 90 days and can be set by the institution anywhere from 7 to 365 days, and since there is no webcam, screen recording or keystroke capture, no footage exists to keep. The homepage lists "GDPR and CCPA compliant by design," "SOC 2 Type II in progress," a public bug bounty, patent pending status, NVIDIA Inception membership, and an advisor who chairs the e-Assessment Association. The Company menu links to a trust and privacy page, a subprocessors list, a security brief and an editorial policy.

What it does not do

The site is explicit about its limits. Physical and environment security is out of scope: Aiseptor "does not verify identity, watch the room, or catch a phone, a paper note, or an in-person accomplice off-camera." For those risks the company recommends pairing it with a live proctor or an identity-verification step.

Pricing and trial

Aiseptor is "Priced per session, not per seat," and the site offers free access: one page promises "5 sessions free, no credit card." You can also book a call with a founder, or claim a free Red Team audit, in which the company's advisory board, which the homepage says includes one of India's most widely recognized ethical hackers, tries to bypass your current assessment flow live on a 30-minute call and then shows what Aiseptor would stop. The site also runs a research hub with an AI cheating statistics report, a threat index and a glossary.

Who it suits

Aiseptor suits assessment platforms, certification bodies and technical hiring teams whose main worry is AI tools running on the candidate's own device, and who want to avoid recording candidates on camera. The per-session pricing and free sessions make it easy to test. The trade-off is its deliberate scope: an exam that also needs ID checks or room monitoring will need a second tool or a live proctor alongside it, and as a company founded in 2025 its track record so far is a small number of pilots.