Topics
belated
AI
Amazon
Image Credits:Matthias Balk/picture alliance / Getty Images
Apps
Biotech & Health
clime
Cloud Computing
Commerce
Crypto
Enterprise
EVs
Fintech
Fundraising
convenience
punt
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
certificate
Social
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
result
Startup Battlefield
StrictlyVC
Podcasts
video
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Google , following on the heel of OpenAI , publish a policy proposalin answer to the Trump presidential term ’s call for a national “ AI Action Plan . ” The tech behemoth endorse weak copyright limitation on AI grooming , as well as “ balanced ” exportation control that “ protect national security system while enabling U.S. exports and ball-shaped business operation . ”
“ The U.S. needs to pursue an active international economical insurance to recommend for American values and patronize AI conception internationally , ” Google wrote in the document . “ For too long , AI policymaking has paid disproportionate attention to the risks , often snub the price that misguided rule can have on innovation , internal competitiveness , and scientific leadership — a dynamic that is set about to shift under the raw Administration . ”
One of Google ’s more controversial recommendations pertain to the use of goods and services of IP - protect material .
Google argues that “ fair exercise and text - and - data excavation exceptions ” are “ vital ” to AI development and AI - pertain scientific innovation . Like OpenAI , the company search to codify the right for it and rivals to prepare on publicly available data point — include copyrighted data — largely without limitation .
“ These exceptions allow for the enjoyment of copyrighted , publicly available material for AI training without significantly impacting rightsholders , ” Google wrote , “ and avoid often extremely irregular , imbalanced , and lengthy negotiation with data holders during model ontogeny or scientific experiment . ”
Google , which hasreportedlytrained anumber of modelson public , copyright datum , isbattlinglawsuitswith data owners who accuse the ship’s company of failing to send word and make up them before doing so . U.S. royal court have yet to decide whether fairish exercise doctrine effectively shields AI developers from IP litigation .
In its AI insurance policy proposal , Google also take effect withcertain exportation controls imposed under the Biden administration , which it says “ may counteract economic competitiveness goals ” by “ levy disproportionate burdens on U.S. cloud overhaul providers . ” That contrasts with statements from Google competition like Microsoft , which in Januarysaid that it was “ confident”it could “ comply full ” with the rules .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Importantly , the exportation rules , which seek to set the availability of advanced AI Saratoga chip in disfavored countries , carve out exemption for trust business seeking prominent cluster of chips .
Elsewhere in its proposal , Google calls for “ long - condition , sustained ” investments in foundational domestic R&D , pushing back against recent federal efforts toreduce spending and eliminate grant awards . The company say the government should release datasets that might be helpful for commercial AI grooming , and apportion funding to “ early - market R&D ” while insure computation and example are “ widely uncommitted ” to scientists and institutions .
Pointing to the disorderly regulative surround create by the U.S. ’ patchwork of DoS AI Pentateuch , Google urged the authorities to pass Union legislation on AI , include a comprehensive privacy and security department framework . Just over two months into 2025,the number of pending AI bank bill in the U.S. has grown to 781 , according to an online tracking puppet .
Google cautions the U.S. government against imposing what it perceives to be onerous obligations around AI systems , like usage financial obligation obligations . In many cases , Google argues , the developer of a model “ has little to no visibility or control ” over how a model is being used and thus should n’t bear responsibility for misuse .
Historically , Google has opposed law of nature like California ’s overcome SB 1047 , whichclearly laid outwhat would found precautions an AI developer should take before secrete a exemplar and in which cases developers might be held liable for example - induced hurt .
“ Even in cases where a developer provide a model directly to deployers , deployers will often be best placed to understand the risks of downstream uses , implement effective risk management , and transmit post - market monitoring and log , ” Google wrote .
Google in its proposal also called disclosure prerequisite like those being contemplated by the EU “ to a fault broad , ” and said the U.S. government activity should counterbalance transparence normal that require “ divulging trade secrets , allow rival to replicate intersection , or compromise interior security by providing a roadmap to adversaries on how to skirt auspices or jailbreak models . ”
A spring up turn of countries and states have passed laws require AI developer to reveal more about how their system mould . California’sAB 2013mandates that companies developing AI organization print a high - level sum-up of the datasets that they used to train their systems . In the EU , to comply with the AI Act once it comes into force , companies will have to supply modeling deployers with elaborate instructions on the operation , limitations , and peril associated with the modelling .
