Nobody built a monitor - Slack Code, the code channel and section 87 BetrVG

Salesforce announced Slack Code on 20 August 2026. Tagging a coding agent from any Slack conversation opens a dedicated code channel in which the conversation, the plan, the code diffs and a live preview sit together, with automatic channel creation and archiving.

Why it mattersThis is the object the analysis is about. If the artefacts did not sit in one attributed, searchable place, the German co-determination question would not arise.

salesforce2026introduci · siliconangl2026salesforc

Slack Code is available at no extra cost on all Slack subscriptions.

Why it mattersA paid enterprise tier routes through procurement, where legal review happens. A feature that is on for every plan reaches teams before anyone reviews it.

computerwor2026new

Salesforce's own announcement names five founding agent partners - Anthropic's Claude, Cognition's Devin, GitHub's Copilot, ChatGPT and Vercel. Computerworld reports four agents at launch: Claude Code, Devin, GitHub Copilot and Vercel.

Why it mattersThe count is attributed rather than asserted because the primary announcement and the trade coverage differ. Nothing in the argument depends on which number is right.

salesforce2026introduci · computerwor2026new

Slack Code agents inherit Slack's built-in security model, permissions and admin controls, and admins can create guardrails preventing non-technical workers from shipping code without engineering review.

Why it mattersThese are the controls a team will point at when asked whether the rollout is governed. They are access controls, and co-determination is not an access-control question.

salesforce2026introduci · computerwor2026new

Section 87(1)(6) of the German Betriebsverfassungsgesetz gives the works council co-determination over the 'Einfuehrung und Anwendung von technischen Einrichtungen, die dazu bestimmt sind, das Verhalten oder die Leistung der Arbeitnehmer zu ueberwachen'. Where no agreement is reached, section 87(2) sends the matter to the Einigungsstelle, whose award replaces the agreement.

Why it mattersThe statutory wording is where the misreading starts, and the Einigungsstelle is why getting it wrong is expensive: the employer cannot simply decide.

dejureorg2026betrvg

In 1 ABR 16/23, decided 16 July 2024, the Bundesarbeitsgericht held that a headset system allowing supervisors to listen in on employee communications is a technische Einrichtung intended to monitor within the meaning of section 87(1)(6) BetrVG, and that its introduction and use is co-determined 'auch dann, wenn die Gespraeche nicht aufgezeichnet oder gespeichert werden'. The employer's subjective monitoring intention is not decisive; objective suitability is.

Why it mattersThis is the correction. It removes both defences an engineering team reaches for: we did not intend to monitor, and we are not storing anything for review.

bundesarbei2024bag

Article 26(7) of the AI Act requires deployers who are employers to inform workers' representatives and the affected workers before putting into service or using a high-risk AI system at the workplace. Article 26 binds deployers of high-risk AI systems.

Why it mattersIt is the clause practitioners reach for, and it is the wrong one today - both because of its risk tier and because informing is not co-determining.

euartificia2026article

Regulation (EU) 2026/1744, in force since 27 July 2026, moved the Annex III high-risk obligations of the AI Act to 2 December 2027.

Why it mattersIt dates the Article 19 and Article 26(6) log duties for Annex III: the runway for deciding which logs are genuinely under your control runs to 2 December 2027.

eurlex2026regulatio

Will McKeon-White of Forrester identified permission inheritance in multiplayer agent settings as a persistent challenge, noting that different inheritance approaches each carry drawbacks.

Why it mattersAn independent analyst naming permissions as the open problem is worth more than the vendor's assurance that permissions are handled.

computerwor2026new