SShortSingh.
Back to feed

Around 45 open-source AI attack frameworks now automate intrusions, a survey finds

0
·6 views

A catalog of roughly 45 open-source projects has found that many automate attacks by pairing a language model with tools it can run and act on. The researchers grouped 32 of them into five families, with the remainder described as forks or variants. Four frameworks include a secret scanner wired to the agent, though none bundles a full pipeline for pulling and hunting secrets from public GitHub. Shodan scans also found dozens of exposed operator consoles, with 11 of 41 fingerprinted frameworks serving a web interface by default. The report ties the main risk to a few heavily forked projects rather than the most popular repositories, and cites the ARTEX platform's takedown in October 2026 as an example of how quickly such tools spread.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Masttro Launches Read-Only MCP Server for Wealth Data Access by AI Tools

Masttro launched a Model Context Protocol server on November 16, 2026, letting eligible family offices, wealth owners, and advisors connect their wealth records to compatible external AI applications. The server is strictly read-only and exposes holdings, transactions, performance, and entity and ownership structures, while offering no write operations. Access is disabled by default and requires explicit client activation, with administrators able to revoke permissions at any time. Masttro states that once data reaches an external AI provider, its processing and protection fall under that provider's arrangements rather than Masttro's. The article, which is partly promotional for a third-party decision-scoring service, argues that per-read decision controls are not part of Masttro's offering.

0
ProgrammingDEV Community ·

Checking Romanian supplier VAT and inactive status via the ANAF API

Under Romanian tax law, buying from a supplier listed as inactive in ANAF's register blocks the deduction of both the expense and its VAT, subject to certain exceptions. A supplier whose VAT registration was cancelled on specified grounds triggers a narrower rule, costing only the VAT deduction from the day after cancellation. Because a supplier can switch to inactive after a contract is signed, a one-time check at onboarding may not be enough. The article describes ANAF's free keyless endpoint, which checks up to 100 company tax IDs per call, and outlines how to monitor changes while avoiding false alerts from errors or transient 'not found' results. The author notes the content is not tax advice and recommends confirming details with an accountant.

0
ProgrammingDEV Community ·

Node.js Reminder Queues: Designing Background Jobs for Safe Retries and Delivery

The article argues that reminder queues in Node.js should be built for at-least-once execution, with idempotent side effects and an explicit terminal state, since a duplicate notification is easier to recover from than a reminder wrongly marked complete. It recommends modeling reminders separately from their attempts, preserving each attempt's history so operators can reconcile outcomes after crashes. Graceful shutdown should stop new claims, let in-flight work finish, renew leases only while progress continues, and reclassify abandoned claims for redelivery. Where a downstream service lacks idempotency keys or receipts, the author suggests an outbox pattern with reconciliation instead of relying on lease-based retries alone. The piece concludes that the queue is only transport, and that the business record must stay explainable after retries and partial failures.

0
ProgrammingDEV Community ·

Two-Hour Talk on How to Think About AI, Recorded by Dmitry Merkulov and Maksim

Dmitry Merkulov of the School of Great Books and a practitioner who builds websites, bots and agents recorded a conversation of nearly two hours on 15 November. They chose to focus less on usage tips and more on how people should approach AI as a way of thinking. The practitioner described how a former hobby became a source of income, while Merkulov brought philosophical and historical context to the discussion. Topics included treating subscriptions as a learning investment, the risk that talk of machine consciousness shifts responsibility away from people, and the "blind archer" approach of running many projects at once. The recording is slated for release on the School of Great Books YouTube channel after editing.