AI Launches: Week of January 20-24, 2026

This week was packed with major AI announcements. From Adobe revolutionizing image editing to the big three racing to dominate healthcare AI, here's everything that happened in the world of artificial intelligence.

Adobe's Object Mask AI: A New Era for Photoshop

Adobe dropped what might be the most significant Photoshop update in years. Object Mask AI uses advanced segmentation models to automatically identify and isolate objects with unprecedented accuracy.

Key Features

• One-click selection: Click any object and get pixel-perfect masks
• Multi-object recognition: Select multiple objects simultaneously
• Edge refinement: Handles hair, fur, and complex edges automatically
• Video support: Works frame-by-frame in Premiere Pro

Why It Matters

Previously, creating accurate masks for complex subjects like hair or transparent objects could take hours. Object Mask AI reduces this to seconds. For content creators and designers, this is a massive productivity boost.

Before: 30+ minutes for complex hair masking
After: 3 seconds with one click

Availability

• Photoshop 2026 (v26.1)
• Adobe Firefly integration
• Available to all Creative Cloud subscribers

The Healthcare AI Race Heats Up

The biggest story this week is the simultaneous push by OpenAI, Anthropic, and Google into healthcare AI. Each company announced major initiatives within days of each other.

ChatGPT Health by OpenAI

OpenAI launched ChatGPT Health, a specialized version of GPT-4 trained on medical literature and clinical data.

Features:

• HIPAA-compliant conversations
• Integration with electronic health records (EHR)
• Symptom analysis with appropriate disclaimers
• Medical literature summarization
• Drug interaction checking

Partnerships:

• Mayo Clinic
• Cleveland Clinic
• Kaiser Permanente

Claude Healthcare by Anthropic

Anthropic responded with Claude Healthcare, emphasizing safety and accuracy in medical contexts.

Differentiators:

• Constitutional AI principles applied to medical advice
• Explicit uncertainty quantification
• Mandatory human oversight triggers
• Multi-language support for underserved communities

Key Quote:

"Healthcare AI must be held to a higher standard. Claude Healthcare will always express uncertainty when it exists and never replace professional medical judgment." - Dario Amodei, CEO

MedGemma by Google

Google's entry leverages their vast medical imaging datasets and DeepMind research.

Strengths:

• Medical image analysis (X-rays, MRIs, CT scans)
• Integration with Google Health ecosystem
• Federated learning for privacy
• Real-time clinical decision support

Healthcare AI Comparison

Gemini 3 Announcement

Google announced Gemini 3, their next-generation multimodal AI model, scheduled for Q2 2026.

Leaked Benchmarks

While official numbers aren't released, leaked benchmarks suggest:

• 2M token context window (up from 1M in Gemini 2)
• Native video understanding up to 4 hours
• Real-time voice mode with emotion detection
• Code generation matching Claude Opus 4.5

What We Know

• Training completed in December 2025
• Using new TPU v6 architecture
• Focus on reasoning and planning capabilities
• Deeper integration with Google Workspace

Industry Reaction

The AI community is cautiously optimistic. Google has struggled to match OpenAI's GPT series in real-world performance, but Gemini 2's improvements were significant. Gemini 3 could finally close the gap.

Sora 2 Restrictions Tighten

OpenAI's video generation model Sora 2 received major usage restrictions this week, following concerns about deepfakes and misinformation.

New Limitations

Why the Restrictions?

Several incidents drove the change:

1. Political deepfakes: Fake campaign videos spreading on social media
2. Celebrity misuse: Non-consensual videos of public figures
3. News manipulation: Generated footage presented as real events

Community Response

The restrictions sparked debate about AI regulation:

Pro-restriction:

• Necessary to prevent harm
• Protects vulnerable individuals
• Maintains public trust in media

Anti-restriction:

• Limits legitimate creative use
• Doesn't address bad actors who'll find workarounds
• Sets precedent for over-regulation

GPT-Image 1.5: Better Consistency

OpenAI quietly released GPT-Image 1.5, an incremental update focused on consistency and character preservation.

Improvements

• Character consistency: Same character across multiple generations
• Style locking: Maintain exact art style across images
• Text rendering: Better legibility and accuracy
• Hand generation: Fewer anatomical errors

Example Use Case

Prompt: "A character named Maya, short black hair,
         green eyes, wearing a red jacket"
 
Generate: Scene 1, Scene 2, Scene 3
Result: Same Maya in all three scenes

This addresses one of the biggest complaints about AI image generation: the inability to maintain consistent characters across a project.

NVIDIA at CES: The Robotics Vision

NVIDIA's CES 2026 keynote focused heavily on robotics and embodied AI. Jensen Huang outlined a vision where AI moves beyond screens into the physical world.

Project GR00T 2.0

The next generation of NVIDIA's humanoid robot foundation model:

• Improved physics understanding: Better object manipulation
• Real-time adaptation: Learns from mistakes instantly
• Multi-robot coordination: Teams of robots working together
• Natural language commands: "Clean the kitchen" actually works

Isaac Lab Expansion

NVIDIA's robotics simulation platform received major updates:

• 10x faster simulation
• Photo-realistic rendering
• Synthetic data generation at scale
• Direct deployment to physical robots

Hardware Announcements

The Big Picture

NVIDIA is positioning itself as the essential infrastructure provider for the robotics industry. As Huang put it:

"Every robot will have an NVIDIA brain. Every factory will have NVIDIA perception. Every autonomous vehicle will run NVIDIA software."

Weekly Summary

Winners

• Adobe: Object Mask AI is genuinely revolutionary
• Healthcare AI: Competition benefits everyone
• NVIDIA: Clear robotics vision and execution

Losers

• Sora 2 Users: Significant capability reduction
• AI Startups: Big tech moving fast into every vertical

Trends to Watch

1. Healthcare AI: Regulatory battles incoming
2. Video AI: More restrictions likely
3. Robotics: 2026 could be the breakout year
4. Multimodal: Everything is becoming multimodal

What's Next

• February: Expected Gemini 3 beta access
• March: Adobe MAX mini-event with more AI features
• Q2 2026: Claude 4 speculation intensifies

Conclusion

This week showed how quickly the AI landscape is evolving. The healthcare AI race alone could reshape an industry worth trillions. Meanwhile, creative tools are becoming more powerful even as restrictions tighten around potentially harmful applications.

The pattern is clear: AI is getting better, more specialized, and more regulated. Companies that can navigate all three dimensions will define the next era of technology.

Stay tuned for next week's roundup.