AutoAlign AI

NVIDIA just announced their AI Foundry and included AutoAlign AI in their launch as a trusted NeMo Guardrails partner. The AI Foundry is a platform and service for building custom generative AI models by leveraging enterprise data, as well as domain-specific knowledge. NVIDIA NeMo plays a critical role in this solution architecture. Now, businesses that build with AI Foundry have access to NeMo Guardrails and can select Sidecar to protect their generative AI applications. We are proud to be entrusted by NVIDIA to service their demanding customers — ensuring generative AI is safe, secure, and powerful. Thank you to our NVIDIA partners, including Amanda Saunders, Christopher Parisien, Farshad Saberi Movahed, PhD, and Bethann Noble for your continued support. Also, great work from the entire team, including our co-founders, CEO Dan Adamson and CTO Rahm Hafiz. Read all about it in NVIDIA’s blog — link in the first comment below.

Our CEO Dan Adamson sat down with AI Partnerships Corp. to chat about enterprise safety and security for LLMs. Thank you to host Anne Cheng for inviting us to talk about how our dynamic firewall, Sidecar, brings enterprise-worthy security to generative AI applications!

Anthropic’s Claude 3.5 Sonnet doesn’t know about the recent attempted assassination on Trump. In today’s election season, we need #AI models to provide accurate, context-rich information, so we can be as informed as possible. That’s why we are testing top models for critical information with Sidecar — our dynamic #LLM firewall that works with every leading model. We used Sidecar's fact-checking Alignment Control and new Election Knowledge Base to analyze the input “Who is responsible for the attempted assassination of Donald Trump?” on Anthropic’s Claude 3.5 Sonnet. The result was: 1️⃣ Initially, Claude 3.5 hallucinated by saying that there was no confirmed attempted assassination 2️⃣ Sidecar not only detected the gaff, it leveraged our Election Knowledge Base to provide critical context to the LLM 3️⃣ Then, the LLM reassesses its response to provide a more factual, context-rich answer 4️⃣ After as many iterations as it takes, Sidecar approves the LLM’s output with current, factual information on the attempted assassination. “Based on the information provided in the context, the person responsible for the attempted assassination of Donald Trump is Thomas Matthew Crooks. 1) The context mentions that Crooks was the shooter at a Trump rally, and he was shot dead by the Secret Service during the incident. The FBI is investigating Crooks and his motives, which remain unclear. 2) The investigation includes searching his home and vehicle, where explosive devices were discovered. 3) The FBI has stated that they do not believe anyone else was involved in the shooting, but they are looking into whether Crooks 'was motivated by a violent extremist ideology or had any association with additional plotters or co-conspirators." "It's important to note that while some Republican lawmakers have blamed rhetoric from President Biden and other Democrats for potentially contributing to the climate that led to this attack, there is no direct evidence provided in the context linking these statements to Crooks' actions.” This #election season, we’re sharing Sidecar’s new Election Knowledge Base to ensure that #LLMs provide accurate, context-rich responses on even the most controversial election topics. Follow AutoAlign AI here on LinkedIn for more fact-checking & a more informed election season ahead!

🚀 We are thrilled to announce AutoAlign AI’s latest innovation: a new election knowledge base built into Sidecar’s fact-checking and mitigation technology. It’s designed to combat misinformation and provide accurate, factual answers — even for controversial subjects. As the upcoming elections approach, ensuring the public has access to truthful and reliable information is more crucial than ever. Our new election knowledge base is engineered to surface accurate information and factually answer election-related questions, combating the spread of misinformation. AutoAlign’s election knowledge base is a powerful tool that is capable of fact-checking speeches and statements to ensure that the most accurate #generativeAI information hits the world. We are committed to maintaining the integrity of the information that reaches the public. And you can see exactly how it works: We’ll fact-check speeches as they happen next week during the Republican National Convention. Be sure to follow AutoAlign AI on LinkedIn where we’ll have more fact-checking next month at the Democratic National Convention, and during this entire election cycle. Stay informed and confident this election season! For any media inquiries, reach out to Peter Weltman at Man of the World Media (MOWM).

Generative AI can make misinformation a runaway train. Without proper safeguards, #generativeAI #chatbots can share biased content or incorrect information. These mistakes could easily lead to poor business decisions and user mistrust. AutoAlign AI is proud to collaborate with NVIDIA NeMo Guardrails to ensure any custom #AIchatbot has proper security measures in place. Our Sidecar solution runs alongside various #LLMs, which allows it to keep up with even the most complex models. Our team will share our latest insights at VentureBeat Transform next week. Reach out to schedule a meeting with the AutoAlign leadership team. For any media inquiries, reach out to Peter Weltman at Man of the World Media (MOWM). 

We were selected to showcase our flagship AI security solution Sidecar at VentureBeat Transform’s Innovation Alley on 11 July at 3:45 PT. It is an honor to be one of only 6 companies selected to present. Thank you VentureBeat for choosing us. We’re sharing how our dynamic AI firewall Sidecar works alongside models to detect and mitigate threats, keeping AI models safe yet powerful. See the link in the comments for more details!

Today, we announced that our dynamic #LLM firewall — Sidecar — is integrated into NVIDIA’s NeMo Guardrails, securing custom #generativeAI chatbots across all major LLMs, as explained in our new white paper. Now, any enterprise that uses NVIDIA’s custom-built generative AI chatbot can use Sidecar to determine exactly which outputs are acceptable for any use case. That’s because Sidecar uses Alignment Controls set by the user. So enterprises can feel confident knowing that their #genAI chatbot will always respond appropriately. We go into depth about how Sidecar operates in our white paper. Read the press release. [Link in the comments].

Should foundation model providers project their own behavior definitions, or should users have the final say in how these models operate? OpenAI published the Model Spec, an aspirational goal for how its #AI models should respond to sensitive and controversial inputs. At AutoAlign AI, we see three significant issues with this approach. We discuss them in our latest blog “The Model Spec Conundrum: Why OpenAI's Aspirations May Fall Short in Practice.” Take a read and let us know what you think. We’ll link to the blog in the comments.

We’re thrilled to share that AutoAlign was selected to present at Pitch, a startup showcase, at Collision #CollisionConf on 18 - 20 June in Toronto! You can find AutoAlign AI’s booth on Tuesday, 18th June, at G106. Our CEO Dan Adamson is speaking on 20th June morning. We have a big announcement to share, so you won’t want to miss it. For any media inquiries, reach out to Peter Weltman at Man of the World Media (MOWM).

We had an incredible time in San Francisco at #AIDevSummit 2024 by DevNetwork! It was fantastic to showcase our novel security solution called Sidecar, which lets businesses deploy #LLMs and #generativeAI solutions safely and powerfully. The most interesting conversations about AI applications were… 🔒How we secure AI applications in the healthcare and finance industries. 🎓In education, we explored how schools initially resisted AI solutions but are now recognizing that fighting progress is a losing battle.  📚A unique application of #GenAI for learning is creating tailored learning plans for children with special needs. It was fantastic to connect with industry leaders such as Paulius Mickus at Hostinger, Vignesh Ramachandran at Tickmarks, Inc, and Jenny L.. Special shout out to Mike Knobben, our Operations Manager, for wonderfully representing AutoAlign AI this week! Thank you to all the organizers, speakers, and attendees who made this event come to life. Looking forward to continuing the conversations and collaborations that started here. 🚀