AI News Daily 2026-10-18
- OpenAI has begun rolling out "ChatGPT Slides," a built-in feature for creating, editing and co-authoring presentation decks inside ChatGPT, with export to PowerPoint and Google Slides, starting with Pro, Business and Enterprise (Space) users.
- Google announced general availability of "Deep Review," an AI tool for analyzing technical documents, at its Research@2026 event, citing use on more than 15,000 conference submissions.
- Anthropic introduced the "Anthropic Cyber Mission," a long-term program to support system security through tools, research and other resources.
- Anthropic began giving the independent evaluator METR employee-equivalent access to its systems, aiming to strengthen the effectiveness of outside safety assessments of frontier models.
- NVIDIA pledged $1 billion in compute resources and support over the next five years to advance US scientific research.
01 OpenAI launches ChatGPT Slides for in-chat deck creation and collaboration
Published: 2026-10-16
Facts. On October 16, 2026, OpenAI began rolling out "ChatGPT Slides," a new capability that lets users create, edit and collaborate on presentation decks directly inside ChatGPT. The feature supports direct editing of text, images and charts, allows teams to co-edit a deck, and can export the result to PowerPoint or Google Slides formats. It is being made available first to Pro, Business and Enterprise users who work in Spaces.
Background. The release extends OpenAI's push to turn ChatGPT into a workspace for producing finished business documents rather than just text and code, following a broader industry pattern of embedding document-creation tools directly into chat interfaces.
Implications. By letting a presentation be drafted, revised and shared without leaving a chat session, the feature could meaningfully shorten the time teams spend assembling proposals and internal materials, and it puts ChatGPT in more direct competition with established office-productivity suites such as Microsoft PowerPoint and Google Slides.
Source: OpenAI — ChatGPT release notes
02 Google brings its "Deep Review" research-analysis tool to general availability
Published: 2026-10-14
Facts. At its Research@2026 event, Google announced that "Deep Review," an AI tool designed to analyze technical documents, is now generally available. Google said the tool has already been used on more than 15,000 submissions at major academic conferences.
Background. The announcement was part of a wider set of research-capability updates Google presented at the Research@2026 event, aimed at R&D and academic audiences that handle large volumes of technical literature.
Implications. General availability of a tool with this scale of prior use suggests literature-review work in R&D organizations can be meaningfully accelerated, continuing a broader trend of AI moving deeper into the day-to-day workflows of research departments.
03 Anthropic revises its 2026 Usage Policy
Published: 2026-10-08
Facts. Anthropic announced a 2026 revision of its Usage Policy for Claude, citing improvements in model capability and feedback from customers as the reasons for the update. The revision updates guidelines covering commercial-use conditions and the scope of prohibited uses.
Background. Usage policies are periodically revised by AI providers as model capabilities and real-world deployment patterns evolve; this is Anthropic's 2026 update to that governing document.
Implications. Because the revision touches commercial-use conditions and prohibited-use scope, organizations deploying Claude in business workflows will need to review the updated terms and may need to adjust internal compliance rules accordingly.
04 Anthropic launches the "Anthropic Cyber Mission" initiative
Published: 2026-10-08 18:04
Facts. Anthropic announced the launch of the "Anthropic Cyber Mission," described as a long-term effort to help secure systems by providing tools, research and other resources.
Background. The initiative sits alongside a broader set of recent Anthropic announcements concerning safety and security, including its policy update and its expansion of external-evaluator access covered elsewhere in this edition.
Implications. The move signals a sustained commitment of resources toward countering AI-enabled cyber threats and gives enterprise security teams a new avenue for engaging with Anthropic's tools and research on this front.
05 Anthropic gives external evaluator METR employee-level access to its systems
Published: 2026-10-15
Facts. Anthropic announced that it has begun giving METR, a nonprofit organization that evaluates frontier AI models, the same level of system access as its own employees. The stated goal is to make METR's independent evaluations more effective.
Background. METR is an external, non-profit evaluator of frontier AI models; granting it employee-equivalent access is a step beyond typical arm's-length audit arrangements.
Implications. Greater transparency toward an independent evaluator could help shape industry norms for external oversight of frontier AI safety, and may put pressure on other AI developers to consider similar arrangements.
06 NVIDIA commits $1 billion to US scientific research over five years
Published: 2026-10-08
Facts. NVIDIA announced it will provide $1 billion worth of compute resources and support over the next five years to advance scientific research in the United States.
Background. The commitment is one of several large, multi-year investments NVIDIA has made recently in basic research and safety work, alongside the UK agentic-safety commitment covered in the next item.
Implications. Investment of this scale from a leading AI-compute provider into basic science points to expanding opportunities for industry-academia collaboration built around AI infrastructure.
Source: NVIDIA — NVIDIA commits $1 billion to advance US science over the next five years
07 NVIDIA commits $250 million to UK agentic-AI safety research
Published: 2026-10-13
Facts. NVIDIA announced it will provide $250 million worth of compute resources over the next five years to support agentic AI safety and security research in the United Kingdom.
Background. The announcement extends NVIDIA's recent pattern of large, multi-year commitments to safety- and science-focused research programs, following its separate $1 billion US science commitment.
Implications. A large-scale investment specifically earmarked for agentic-AI safety research is likely to influence how safety standards for autonomous AI systems are developed as enterprises adopt agentic tools more widely.
08 Adobe and NVIDIA expand their agentic creative-tools collaboration
Published: 2026-10-15
Facts. Adobe and NVIDIA announced an expansion of their collaboration aimed at speeding up agentic creative production, with optimization work spanning core Adobe products including Photoshop, Premiere, Substance 3D, Frame.io and Brand Concierge.
Background. The expanded partnership builds on the two companies' existing relationship around AI-accelerated creative software.
Implications. Deeper integration of agentic AI across Adobe's flagship creative applications could translate into productivity gains for design and marketing teams that rely on these tools day to day.
Source: NVIDIA — Adobe and NVIDIA expand collaboration to speed agentic creation
09 Editor's note
Today's items split into two connected threads. One is a continued build-out of AI inside everyday office and creative workflows: OpenAI's ChatGPT Slides and Google's now-generally-available Deep Review both move document creation and technical-literature analysis directly into chat-based tools, while Adobe and NVIDIA's expanded collaboration pushes agentic AI deeper into established creative software such as Photoshop and Premiere. The other thread is governance and trust: Anthropic's revised Usage Policy, its new Cyber Mission initiative, and its decision to give the external evaluator METR employee-equivalent system access all point toward major providers pairing capability growth with explicit commitments on safety, security and independent oversight. NVIDIA's five-year, multi-hundred-million-dollar commitments to US scientific research and UK agentic-safety research sit at the intersection of both threads, funding the compute that underpins new capabilities while also targeting safety-specific research.
On process: this edition is a mix of same-window items and selected retrospective items from early-to-mid October, used to fill out the day's coverage after the search budget for new items was exhausted. A small number of other candidate items considered for today's window — including a possible regulatory development and a company valuation report — were held back because only a single source could be confirmed before the search budget ran out; our sourcing standard requires independent confirmation from a second outlet before publication, and none of the items above required that trade-off.