GitHub has announced the general availability of AI Scan for pull requests and Agentic Autofix for code scanning alerts.[1][2] These features bring AI assistance closer to the work of securing code. They do, however, operate on different scopes: proposed changes for the former, alerts on the default branch for the latter. For businesses, the challenge is to prepare fixes more effectively while keeping technical validation in place before they are merged.
AI Scan reviews changes before they are merged
A pull request is used to propose a change and have it reviewed before it is merged. AI Scan complements CodeQL there, notably for PHP, Bash, Terraform and certain less well-covered frameworks. Once enabled alongside code scanning, it surfaces findings flagged as AI-generated, sometimes accompanied by an explanation and a suggested fix.[1]
These findings remain advisory: they do not cause the code scanning check to fail and do not block merging via ruleset requirements. AI Scan does not process pull requests from forks or those created by Dependabot.[1] A team must therefore organise how these findings are read and dealt with, without assuming that their presence automatically holds up a release.
Agentic Autofix prepares a fix from an existing alert
Agentic Autofix can be assigned a code scanning alert located on the default branch. It explores the relevant files, proposes a change, re-runs CodeQL and can iterate before opening a draft pull request for review.[2]
GitHub explicitly excludes alerts that exist only in a pull request or another branch from this assignment to Copilot.[2] The two new features should therefore not be described as an automatic pipeline in which every AI Scan finding is immediately handed over to Agentic Autofix.
An alert disappearing on a new scan gives a useful indication about the targeted issue. It does not prove that the application retains all its expected behaviour, nor that the change introduces no other defect. Functional tests and a review of the change remain necessary.
Two paths and a human decision
The table distinguishes between the announced features and the responsibility of the team using them.
| Stage | What GitHub provides | Team check |
|---|---|---|
| Detection during review | AI Scan flags risks in proposed changes.[1] | Assess the risk and decide how to handle it before merging. |
| Preparing the fix | Agentic Autofix works on an alert from the default branch.[2] | Review the modified files and the relevance of the fix. |
| Validation | A new CodeQL scan accompanies the agent’s work.[2] | Also check business tests and side effects. |
Access conditions matter when choosing a pilot
AI Scan is available for public repositories; private or internal organisation repositories require GitHub Code Security or GitHub Advanced Security. A Copilot licence is not required.[1] Agentic Autofix is aimed at customers of these offerings on GitHub Enterprise Cloud, with administrative activation and a defined Copilot billing method.[2]
Both features consume Code Security AI credits; Agentic Autofix also uses GitHub Actions minutes.[1][2] Before rolling out, SMEs need to check their plan, the scope of repositories and the associated budget. The economic benefit will depend as much on review time as on the number of fixes actually accepted.
SMEs can evaluate results in two distinct situations
First illustrative scenario: a team modifies a PHP form and receives a finding in its pull request. The developer checks whether user input can actually reach the sensitive operation, examines the suggestion, then adds a test reproducing the case. A plausible answer from the AI is not enough to establish whether the risk exists or has gone away.
Second scenario: a code scanning alert already exists on the default branch of an eligible repository. The team assigns it to Agentic Autofix, then reviews the proposal. It checks that the fix does not inadvertently change access rights, accepted formats or business results. These scenarios are an evaluation approach, not published customer results.
The pilot can track the time to a validated fix, the proportion of suggestions retained and the regressions detected. Counting only closed alerts could mask an excessive verification workload or unsuitable changes.
Developers must understand the vulnerability as well as the fix
This development raises the bar for Web Programming and Management. Reading code remains essential to understand the path data takes, spot a fragile assumption and determine whether the fix addresses the root cause. Professionals must be able to explain why a change reduces the risk.
The skill is also reflected in the quality of tests. A case that fails before the fix and passes afterwards provides a useful benchmark, to be complemented by checks on legitimate uses. In particular, a stricter control must not prevent a necessary function from working or shift the vulnerability elsewhere.
Finally, the developer organises the review and documents decisions. They must know when to accept a suggestion, modify it or reject it. Delegating to an agent can speed up preparatory work, but command of the application remains essential to decide what gets merged.
Editor’s note
The progress to measure is a fix that is understood, tested and maintainable. These tools can help reduce research and preparation work; their real value will depend on the coverage of the scans and on teams’ ability to verify the proposals.
Further reading
Four CCCLX reads on assisted development and software security.
Sandbox, terminal, policies: GitHub puts guardrails on its code agents
Project Glasswing: Anthropic goes on the offensive against software vulnerabilities
Open-source security changes scale: IBM, Red Hat and Palo Alto join forces
With Code Layers, Figma brings design and development firmly together
Sources
- 01
GitHub Changelog (15 October 2026). “AI Scan for pull requests generally available”. Overview of the general availability, scope and terms of use of AI Scan.
- 02
GitHub Changelog (15 October 2026). “Agentic autofix for code scanning alerts general availability”. Details of how Agentic Autofix works, its access conditions and billing.
Mistral Large 4: what its trillion-parameter model could change for businesses
Mistral has opened the public preview of Mistral Large 4, nicknamed “Le Chonk”, and plans to release its weights by the end of the month.[1] This European model targets agentic use cases in particular, as well as cybersecurity, finance and industry. For businesses,...
IBM Bob: developing with AI without your code leaving the company
IIBM has announced a self-hosted version of Bob, its agentic development platform, for internal infrastructure, private or sovereign clouds and environments isolated from the internet.[1] This option addresses a concrete obstacle: some applications contain code, data...
HPE is preparing a network that anticipates incidents and automates certain interventions
HPE plans to deploy a Self-Driving Network for the 2027 Ryder Cup, with Wi-Fi 7 and an operations centre bringing together data and AI tools.[1] The stated aim is to monitor usage, adapt the network and move towards autonomous interventions. For businesses, the...
Can your AI agent really defend your company? Hack The Box puts its skills to the test
Hack The Box is launching AI Range Enterprise Edition to let security teams assess their own AI agents on the roles assigned to them.[1] For businesses, the stakes are concrete: having evidence before delegating a task, then checking that this evidence remains valid...
VAST Data wants AI to work with sensitive data in a protected environment
VAST Data has unveiled DataEnclave, a confidential execution capability designed to bring artificial intelligence models closer to companies’ sensitive data. Presented as a preview, it is due to go on sale in the first quarter of 2027.[1] The aim is to make certain...
Databricks buys Row Zero to bring spreadsheets closer to governed data
Databricks has announced the acquisition of Row Zero and plans to integrate its spreadsheet experience into Genie.[1] The aim is to let business teams explore data in a familiar interface, with greater control over access and usage. For companies, the challenge is to...
Google Cloud takes its X5 instances to 48 TB for large SAP databases
Google Cloud has announced the general availability of the X5 series, which raises the maximum advertised memory capacity on a single node to 48 TB, compared with 32 TB for X4.[1] Aimed in particular at very large SAP HANA databases, the offering is based on bare...
Cloud Storage: Google automates the detection of anomalies
A rise in reads, cross-region transfers or stored volume can inflate the cloud bill and disrupt applications. With Storage Intelligence advisor, Google offers a view of unusual changes in Cloud Storage, along with leads for investigating them. For businesses, the...
AI voice agents: Oracle details a LiveKit architecture on OCI
Oracle has described an architecture that pairs the LiveKit Agents framework with Oracle Cloud Infrastructure to automate real-time voice calls. The design links telephony, AI models and business services, while placing internal components in private subnets. For...
GitHub Copilot bulks up for business: coding agents come under IT control
Copilot is no longer just there to suggest the next line GitHub Copilot is continuing its transformation. Originally known mainly for completing code in the editor, the tool is now evolving towards agents capable of exploring a project, modifying files, using tools...
