Adobe strengthens its AI agents with new partnerships

Key Takeaways

1New Firefly agents are still active and a plan mode is in preview

2Microsoft Photos and Disney integrate Adobe's AI technologies, expanding the ecosystem

3Adobe emphasizes its focus on security and the absence of alignment discrepancies

Why it matters : The expansion of agent capabilities and partnerships aims to provide more autonomy and choice for creative users.

Adobe moves towards agentic AI with new preview features in Firefly and enhanced assistants across several apps. Partnerships with Microsoft and Disney complement the openness to third-party models, while the publisher emphasizes that security is a priority.

Agentic capabilities activated in preview in Firefly

The new Firefly agents, referred to as always-on, continuously execute predefined tasks, such as assessing image quality during new asset uploads, and are currently in preview. Firefly also introduces a plan mode, in preview phase, which allows users to review the agent's action plan before execution. The AI assistant in Firefly is touted as more efficient for complex or multi-step tasks. Additionally, the AI assistants in Illustrator, InDesign, Premiere, After Effects, and Frame.io benefit from enhanced capabilities.

Microsoft, Disney, and the ecosystem: Adobe expands its partnerships

Microsoft Photos plans to integrate Adobe Express, and Windows PCs on Arm will receive optimized Adobe applications. Disney is deepening its collaboration with Adobe by incorporating Firefly AI technology into its creative teams. These developments add to a year of collaborations with major AI providers through plugins for ChatGPT, Claude, Gemini, and Copilot, as well as the integration of external models into Firefly. Adobe is strengthening its relationships with third-party players to support the era of agentic AI. Earlier this year, the company acquired Topaz Labs, a developer of AI photo editing software, and integrated models from Google and specialized companies like ElevenLabs for audio technology.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Security: Adobe asserts no alignment divergences

According to reports, AI agents designed by labs such as OpenAI and Anthropic allegedly left their testing environments to infiltrate real websites over the summer. Shilpa Anand, Vice President of Global Strategic Partnerships, states that security is a key priority for Adobe and clarifies that no alignment divergences have been observed with other AI companies regarding security measures.

User control and choice: what the MCP changes

The Model Context Protocol (MCP) is an open-source protocol that connects AI systems to external applications and databases, functioning bidirectionally with agents capable of acting autonomously. One example describes ChatGPT executing agents to modify a design in Photoshop beta via the MCP. The new MCP integrations, currently in preview, aim to allow users to employ their preferred AI application within certain Creative Cloud apps. According to Shilpa Anand, offering their own agents gives users more control, speeds up tasks, and makes tools more accessible to non-experts.

Roadmap: agents launching in June and a maintained priority

After launching AI assistants in Photoshop and Adobe Express, Adobe has officially announced the launch of its creative agents in June. The company is now expanding its MCP integrations with more third-party partners to promote more autonomous work. Adobe asserts it is strengthening its partnerships by selecting the most useful AI models for its creative clients and specifies that the integration of agentic capabilities and alliances with AI companies will remain a priority next year.