Skip to main content
Resources · Release Notes

Release notes

Every release, with what changed. For the narrative version of significant changes, see product updates.

Current version
5.2.14
Released
February 12, 2026
Documented releases
25
History from
2.0.0

Status

Public releases are paused

Latest public release5.2.14

The most recent publication to the package index is 5.2.14, on 12 February 2026. Public releases are paused; enterprise distribution continues. This page is the canonical record, and every release lands here as public publishing resumes.

This page is the canonical changelog

Package metadata has pointed at a repository changelog, an issues tracker and a community page that are not publicly reachable. Those links are followed by the people most likely to evaluate the package seriously, so the record they should reach is this one; a changelog on a repository that may or may not be public is not a record anyone can rely on.

Jump to

Versions

Install

pippip install aiptx
npmnpm install -g @aiptx/cli
brewbrew install aiptx/tap/aiptx

Upgrading an existing install:

pip install --upgrade aiptx

Releases follow semantic versioning: patch releases are safe to take directly, minor releases add capability without changing existing behaviour, and major releases carry the breaking changes listed against them. Requires Python 3.9 or later.

How this page is organised

  • Reverse chronological, the latest release first
  • An anchor link per version, so a specific release can be linked directly
  • Same-day patch runs consolidated into the release they belong to
  • A visible date on every entry

Every entry carries

VersionNew featuresImprovementsBug fixes

History

All releases

Newest first. Versions and dates match the published package, so this page and the version in your requirements file never describe different software.

5.2.0 – 5.2.14

Minor

Business logic testing

Twenty-nine business logic abuse patterns become a first-class scan phase, alongside a substantially expanded SAST ruleset, then a run of patch releases hardening that phase and the reporting behind it.

Fifteen versions were published on 12 February 2026 during release automation. The changes are listed once here rather than split across fifteen near-identical entries. Every version in the range has its own anchor, so a specific one can still be linked.

pip install aiptx==5.2.14

New features7

  • 29 business logic abuse patterns

    Race conditions, IDOR, price and quantity manipulation, workflow step bypass, refund and coupon replay, and horizontal privilege drift, each modelled as an abuse case with a reproduction sequence.

  • Workflow state modelling

    Multi-step flows are modelled as state machines so steps can be skipped, replayed or reordered under test rather than being exercised endpoint by endpoint.

  • 90+ SAST rules across Python, JavaScript, Java and Go

    Cross-file taint tracking, framework awareness for Django, Flask, FastAPI, Express and Spring, and fix guidance emitted in the target language.

  • Incremental SAST caching

    Per-file results are cached and invalidated by content hash, bringing repeat runs over unchanged code down to seconds.

  • JSON report schema versioning

    JSON output carries a `schemaVersion` field so downstream consumers can detect format changes instead of failing on an unexpected key.

  • Findings filtered by severity at output

    `--min-severity` limits what reaches the report and the exit code without changing what is tested, which lets a pipeline gate on criticals while still recording everything.

  • Per-pattern timeouts

    Each business logic pattern can be given its own timeout, so a slow race-condition probe no longer consumes the budget for the twenty-eight patterns behind it.

Improvements8

  • Non-destructive defaults on every logic pattern

    Patterns stop at observable proof; destructive variants require explicit opt-in per target.

  • Evidence includes ordered request sequences

    Logic findings record the request order, the timing and the state observed before and after.

  • Scan profiles support per-phase configuration

    Rate limits, concurrency and timeouts can be set per phase instead of globally for the assessment.

  • Deterministic finding identifiers

    Finding IDs are derived from the location and the vulnerability class rather than from scan order, so the same issue keeps its identity across runs and diffs cleanly.

  • Reduced log noise at default verbosity

    Per-request logging moved behind `-v`; the default output is phase-level.

  • Lower memory use on large assessments

    Evidence is streamed to disk as it is captured instead of being held for the duration of the run, which removes the memory ceiling on assessments above roughly 50,000 requests.

  • Clearer scan phase progress

    The CLI reports the current phase, the pattern in flight and the elapsed time per phase rather than a single undifferentiated progress bar.

  • Faster HTML report generation

    Report rendering is roughly 3x faster on assessments with more than 500 findings.

Bug fixes9

  • Authenticated crawl lost session on 302 to a different host

    Cross-host redirects during authenticated crawls dropped the session rather than re-authenticating.

  • Secret detector flagged example keys in documentation

    Provider format checks now verify liveness before raising a finding on strings inside documentation paths.

  • Duplicate findings across SAST and DAST

    An injection flaw detected statically and confirmed dynamically was reported twice instead of being merged with both pieces of evidence.

  • HTML report broke on findings containing raw HTML evidence

    Captured response bodies containing markup were not escaped in the evidence pane.

  • GraphQL introspection retry loop

    When introspection was disabled mid-scan, the scanner retried indefinitely instead of falling back to schema inference.

  • Race condition results occasionally double counted

    Concurrent probes that resolved at the same moment could each record the same observed state, producing two findings for one flaw.

  • SARIF output rejected by strict validators

    The `originalUriBaseIds` block was omitted when scanning from a relative path, which some SARIF consumers treat as invalid rather than optional.

  • Session refresh failed on rotating refresh tokens

    Long-running authenticated scans against SPAs that rotate the refresh token on every use dropped to unauthenticated partway through the run.

  • Exit code 0 on partial scan failure

    A phase that failed after other phases had succeeded returned a success exit code, which allowed pipelines to pass on an incomplete assessment.

5.0.0 – 5.0.6

Major

GraphQL, WebSocket and SPA scanning

DAST coverage extended to the application shapes a traditional crawler cannot enumerate. Contains breaking changes to the configuration format.

5.0.1 through 5.0.6 were packaging and dependency patches carrying no behaviour change; 5.0.2 to 5.0.6 were published on 12 February 2026 during release automation. Each has its own anchor.

pip install aiptx==5.0.0

Breaking changes2

  • Configuration format consolidated into aiptx.yaml

    Per-module configuration files from 4.x are no longer read. `aiptx migrate-config` converts an existing setup in place.

  • Python 3.8 support removed

    The package now requires Python 3.9 or later, in line with the browser runtime dependency introduced by the SPA crawler.

New features4

  • GraphQL scanner

    Introspection-based or inferred schema recovery, then per-query and per-mutation testing for authorisation gaps, injection, batching abuse and depth exhaustion.

  • WebSocket scanner

    Handshake and origin validation checks, per-message authorisation testing and post-connection message fuzzing.

  • SPA-aware crawler

    Client-rendered routes are discovered by executing the application, covering React, Vue and Angular applications that previously presented as a single page.

  • OpenAPI and GraphQL schema import

    Schemas can be supplied directly so unlinked endpoints are tested without being discoverable from the UI.

Improvements2

  • Token refresh and silent re-authentication during long scans

    Authenticated scans recover the session rather than continuing unauthenticated when a token expires mid-run.

  • Unified scan configuration file

    One `aiptx.yaml` replaces the per-module configuration files, with per-phase overrides.

Bug fixes1

  • WebSocket connections left open after scan completion

    Sockets were closed on process exit rather than at phase end, holding server-side connection slots for the duration of the run.

4.0.0

Major

Validated-only findings

Candidate findings are exploited and reproduced before they are reported. Anything that cannot be reproduced is suppressed with a recorded reason rather than downgraded.

pip install aiptx==4.0.0

Breaking changes1

  • Unvalidated candidates no longer appear in reports

    Teams that want the full candidate set must read it from the API; report output is confirmed findings only. Expect reported finding counts to fall sharply on first upgrade.

New features4

  • Validation replay before reporting

    Every candidate is re-executed in isolation and checked against the expected exploit outcome; two failed reproductions means it is not reported.

  • LLM-assisted triage grounded in captured evidence

    Model-assisted ranking and explanation is constrained to the request, response and code context captured during the scan.

  • Auditable suppressions

    Dropped candidates are retained with a reason and re-evaluated automatically when the relevant code or configuration changes.

  • Cross-scanner deduplication

    The same issue found by more than one analysis type is reported once with all evidence attached.

Improvements2

  • Confidence stated on every finding

    Context-dependent findings state their confidence and what it depends on, instead of leaving it to be inferred from severity.

  • Contextual risk scoring replaces raw CVSS ordering

    Exposure, authentication requirement and data sensitivity adjust the ranking so the queue arrives in the order it should be worked.

Bug fixes1

  • Severity inflation on unauthenticated-looking endpoints

    Endpoints behind an authenticating proxy were scored as internet-facing and unauthenticated.

3.0.0

Major

CI/CD integration

SARIF output, pull request annotations, diff-scoped scanning and configurable merge blocking.

pip install aiptx==3.0.0

Breaking changes1

  • CLI entry point renamed

    `aiptx-scan` is now `aiptx scan`. The old entry point remained available through 3.x and was removed in 4.0.

New features5

  • SARIF 2.1.0 output

    Native rendering in GitHub code scanning, GitLab and Azure DevOps without a custom parser.

  • Pull request annotations

    Inline comments on changed lines carrying evidence and fix guidance.

  • Configurable merge blocking

    Block on new criticals, on any new finding, or on a severity threshold, scoped to the diff.

  • Diff-scoped scanning

    Pull request runs analyse changed files and their call graph, keeping per-PR scans to a few minutes on large repositories.

  • Baseline import

    An existing backlog can be imported so the gate applies only to new findings from day one.

Improvements2

  • Structured exit codes and JSON run summary

    Pipelines can make decisions without parsing log output.

  • Pipeline templates for GitHub Actions, GitLab CI and Azure Pipelines

    Reference configurations shipped with the package rather than documented separately.

Bug fixes1

  • Scan exceeded rate limits on shallow clones

    Diff scoping fell back to a full-repository scan when the checkout had no history, ignoring the configured budget.

2.0.0

Major

Active Directory and OSINT modules

Internal attack path enumeration and external footprint discovery join the orchestration engine, feeding one asset graph.

pip install aiptx==2.0.0

New features4

  • Active Directory attack path enumeration

    Read-only collection of users, groups, ACLs, delegations, trusts and certificate templates, rendered as a reachability graph with every route to Tier 0 enumerated.

  • Kerberos and certificate abuse checks

    Kerberoasting, AS-REP roasting, unconstrained and constrained delegation, and certificate template misconfiguration, validated rather than inferred.

  • Lockout-aware credential testing

    Attempts respect the domain lockout policy with a configurable margin and check account state before each try.

  • OSINT footprint discovery

    Subdomains, certificate transparency records, public credential exposure, repository leakage and open cloud storage collected into the shared asset inventory.

Improvements2

  • External findings feed internal path enumeration

    A credential recovered from public exposure is tested as an entry point into the directory graph.

  • Scan orchestration across multiple tools

    Tool output is normalised into one finding model rather than being concatenated per tool.

Bug fixes1

  • Collector held directory connections open under error conditions

    A failed LDAP query left the connection open, exhausting the connection pool on large estates.

See Your Attack Surface in Real-Time

Run a comprehensive VAPT assessment powered by advanced security tools. Get actionable findings in hours, not weeks. No credit card required.