Anthropic Launches Claude Tag for Slack and Joins $915 Million Carbon Removal Purchasing Group

Claude Tag: A Shared AI Agent, Not Another Chatbot

Anthropic launched Claude Tag in beta on Tuesday, June 17, replacing its existing Claude in Slack app with something structurally different from prior AI integrations. According to VentureBeat, the product runs on Claude Opus 4, the model Anthropic released less than a month ago.

The core design is "multiplayer." One Claude instance lives inside a Slack channel, interacts with every person in that channel, and accumulates shared context across all of those conversations. Anyone who types @Claude can hand it a task. Anyone else in the channel can see what it is working on and pick up where the last person left off. This represents a departure from single-user AI tools bolted onto messaging platforms.

Claude Tag is available today in beta for Claude Enterprise and Team customers, according to ZDNET. An administrator pairs it with a workspace, defines which channels it can operate in, grants it access to specific tools and data sources, and sets spending limits. From there, the product can run data analyses, write code, pull sales numbers, and break multi-step tasks into stages it executes asynchronously.

If an admin enables "ambient" behavior, Claude Tag watches the full channel, learns from its activity over time, and proactively surfaces information from elsewhere in the organization when it judges that information relevant. Anthropic says it will NOT pull from private channels, and that access to sensitive data "can be very tightly controlled," per Engadget.

The Productivity Claim

Anthropic made a striking assertion in the launch materials: 65% of its own product team's code is now generated by an internal version of Claude Tag. Both VentureBeat and ZDNET reported the figure. That number comes from Anthropic itself, not from an independent audit, so it should be treated as a self-reported marketing claim until a third party verifies it.

The skeptical case is straightforward. A company selling an AI agent has every financial incentive to report high internal adoption numbers. "65% of code" also does not specify what percentage of that code was reviewed, revised, or rejected by human engineers before shipping. The figure tells you something, but not everything.

The claim is specific and falsifiable. If Anthropic's own engineering output degrades publicly over the next year, the number will be revisited. For now, enterprise buyers evaluating the product should treat it as a directional data point, not a warranty.

Governance Risks Are Real

Enterprise technology leaders have a legitimate concern about ambient AI agents accumulating institutional knowledge over time in a shared, semi-public chat environment. An AI that "sets tasks for itself" and monitors channel activity continuously is a fundamentally different governance challenge than a search tool or a one-off coding assistant.

Anthropic's answer is administrator control: admins specify which tools and information Claude Tag can access per channel. Whether that control layer is robust enough for regulated industries, financial services, or legal teams remains an open question the company has not fully answered. The product is in beta. Governance documentation for enterprise compliance use cases has not been independently reviewed as of June 23, 2026.

Frontier Climate: Anthropic Joins a $1 Billion Carbon Removal Bet

Separately, Anthropic joined Frontier Climate's advance market commitment to purchase carbon removals. Frontier announced this on June 17, according to Utility Dive. The group, which also includes JPMorgan Chase, H&M, and Salesforce, launched in 2022 with a target of buying $1 billion in carbon removals by 2030.

Frontier's latest announcement included $915 million in new financing commitments. As of June 17, the group had contracted over 1.8 million tons of carbon removals across 53 projects, with $698 million contracted in total. Since June 2025 alone, Frontier contracted 463,700 tons for $169.8 million.

Tim Thomson, president and CFO of Charm Industrial, which has a $53 million contract with Frontier for 112,003 tons of biomass carbon removal, made the connection between AI and carbon removal explicit. "AI has the potential for incredible economic growth and scientific breakthroughs, but comes at the cost of enormous power requirements," Thomson told ESG Dive. He argued carbon removal can help offset those emissions, including fine particulate matter and nitrogen oxide in addition to CO2.

Frontier said it will narrow its supply-side focus going forward, concentrating on technologies it believes have gigaton-scale potential. The group projects that pathways including direct air capture, ocean alkalinity enhancement, enhanced rock weathering, and biomass storage could reach annual gigaton scale at between $60 and $300 per ton.

The Unresolved Question at the Center of Both Moves

Both announcements reflect the same underlying tension: AI infrastructure, including Claude's own compute demands, generates substantial energy consumption and associated emissions. Anthropic joining a carbon removal purchasing group is an acknowledgment of that reality, not a solution to it. Whether purchased carbon removals are a credible offset for data center emissions, or a way to purchase reputational cover without reducing the actual footprint, is a debate Frontier's own buyers have not settled. Frontier's pivot toward narrower, higher-conviction technology bets suggests even committed buyers are still working out which removal pathways will actually scale.